Escolar Documentos
Profissional Documentos
Cultura Documentos
Quadcopter
Doctor of Philosophy
Alireza Nemati
Submitted to the Graduate Faculty as partial fulfillment of the requirements for the
Doctor of Philosophy Degree in Electrical Engineering
University of Cincinnati
March 2016
The aim of the present work is to model, design, control, fabricate and experimentally study
quadcopter with tilting propellers. A tilting quadcopter is an aerial vehicle whose rotors can
tilt along axes perpendicular to their respective axes of rotation. The tilting rotor quadcopter
provides the added advantage in terms of additional stable configurations, made possible by
additional actuated controls, as compared to a traditional quadcopter without titling rotors. The
tilting rotor quadcopter design is accomplished by using an additional motor for each rotor that
enables the rotor to rotate along the axis of the quadcopter arm.
robots which cannot achieve any arbitrary desired state or configuration. For example, the ve-
hicle cannot hover at a defined point at a tilted angle. It needs to be completely horizontal in
order to hover. An attempt to achieve any pitch or roll angle would result in forward (pitch) mo-
tion or lateral (roll) motion. This proposed tilting rotor concept turns the traditional quadcopter
into an over-actuated flying vehicle allowing us to have complete control over its position and
orientation.
In this work, a dynamic model of the tilting rotor quadcopter vehicle is derived for fly-
ing and hovering modes. The model includes the relationship between vehicle orientation an-
gles and rotor tilt-angles. Furthermore, linear and nonlinear controllers have been designed to
achieve the hovering and navigation capability while having any desired pitch and/or roll orien-
tation. In the linear approach, the four independent speeds of the propellers and their rotations
about the axes of quadcopter arms have been considered as inputs. In order to start tracking a
desired trajectory, first, hovering from the initial starting point is needed. Then, the orientation
ii
of the vehicle to the desired pitch or roll angle is obtained. Subsequently, any further change in
pitch or roll angles, obtained using a linear controller, result in motion of the quadcopter along
The dissertation then presents a nonlinear strategy for controlling the motion of the quad-
copter. The overall control architecture is divided into two sub-controllers. The first controller
is based on the feedback linearization control derived from the dynamic model of the tilting
quadcopter. This controls the pitch, roll, and yaw motions required for movement along an ar-
bitrary trajectory in space. The second controller is based on two Proportional Derivative (PD)
controllers which are used to control the tilting of the quadcopter independently along the pitch
and the yaw directions respectively. The overall control enables the quadcopter to combine
Furthermore, the stability and control of tilting-rotor quadcopter is presented upon failure
of one propeller during flight. On failure of one propeller, the quadcopter has a tendency of
spinning about the primary axis fixed to the vehicle as an outcome of the asymmetry about the
and it is capable of handling a propeller failure, thus making it robust in one propeller fail-
ure during the flight. The dynamics of the vehicle once the failure accrued is derived and a
The dynamic model and the controller of the vehicle were verified with the help of numer-
ical studies for different flight scenarios as well as failure mode. Subsequently, two different
models of the vehicle were designed, fabricated and tested. Experimental results have validated
iii
Copyright 2016, Alireza Nemati
This document is copyrighted material. Under copyright law, no parts of this document may
be reproduced without the expressed permission of the author.
iv
Acknowledgments
First and foremost, I would like to express my sincere gratitude to my advisor Professor
Manish Kumar for the continuous support of my Ph.D. study and related research, for his
patience, motivation, generosity, enthusiasm and immense knowledge. I appreciate all his con-
tribution of time, ideas, encouragement and financial support that he provided. His continuous
guidance helped me carry out research and write this thesis. During my study, he was not only
a great advisor for problems related to my work, but also an exceptional consultant for other
situations I faced.
I wish to express my sincere thanks to Professor Ali Minai, my thesis advisor, for his
valuable guidance and encouragement extended to me. I am thankful to Dr. Kelly Cohen for
all his motivations he gave not only to me but also to the other members of our team, for his
Besides my advisors, I would like to thank Professor Raj Bhatnagar, and Professor Rui Dai,
I addition, a thank you to Younes Kheradmand, who was my boss before I started my Ph.D.
for 5 years. I really appreciate all the advise he provided me like a father and encouragements
he gave like a friend during my career. A thank you also to professor Abdullah afjeh who has
My appreciation also extends to my fellow lab mates. Balaji R Sharma, Baisravan Hom-
chaudhuri and Ruoyu Tan who helped me settle in the lab for first couple of months. I would
like to thank my friends at the Cooperative Distributed Systems Laboratory for the wonderful
times, long meetings and exciting brainstorming discussions we had. Two of my colleagues
cannot remain nameless, Mohammad Sarim and Mohammadreza Radmanesh for the stimu-
lating discussions, for the sleepless nights we were working together before competitions and
deadlines, for the travels we made for different conferences and for all the fun we had in the
v
last couple years.
The final words in acknowledgments are usually reserved to those dearest to the author. I
do not wish to break this tradition. Without the love and support from my family, I would not
vi
Table of Contents
Abstract iii
Acknowledgments v
Contents vii
List of Tables x
List of Figures xi
1 Introduction 1
1.1 Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Approaches . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.4 Contribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.5 Publications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
vii
3 Dynamic Modeling 27
4 Control System 35
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
6 Hardware design 70
6.2.1 Motors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
6.2.2 Batteries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
6.3 Avionics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
6.3.2 Communications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
viii
7 Numerical Simulations and Experimental Results 80
7.2.2 Trajectory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
7.2.2.1 Line . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
7.2.2.2 Box . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
References 104
ix
List of Tables
x
List of Figures
2-1 Gyroplane. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2-3 De Bothezat . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2-5 George Lehbergers 1930 tilting propeller vertical take-off “flying machine”. . . . . 13
2-14 The Configuration of the QTR UAV under research a) Quadcopter mode b)Flying
2-17 CAD model of the quadcopter with tilting propellers, Max Planck Institute. . . . . 25
2-18 Vehicle prototype on the ball joint rig flight test, Cranfield University. . . . . . . . 26
3-1 Schematic diagram showing the coordinate systems and forces acting on the quad-
rotor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
xi
4-2 Hovering with tilted arms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
6-6 The CAD model of tilting mechanism mounted on the arm with the servo and the
motor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
7-8 Position and orientation of the quadcopter: altitude vs. time (top left), x-position
vs. y-position (top right), pitch vs. time (bottom left), and yaw vs. time (bottom
right). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
7-9 The reference roll (bottom) and actual roll angle (top) during the flight. . . . . . . 86
xii
7-14 The actual orientation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
7-22 The actual orientation of the vehicle along the three directions . . . . . . . . . . . 95
xiii
List of Symbols
xiv
Chapter 1
Introduction
This chapter discusses the objectives, approaches, applications and contributions of the
research. The list of publications resulting from the present work are also provided.
1.1 Objectives
The primary objective of this work is to model, design, control, fabricate and experimen-
tally study quadcopter with tilting propellers. The tilting propellers are expected to result in
more stability as well as the ability to follow any arbitrary trajectory in a smoother manner as
compared to the conventional quadcopters and also add the capability to continue the mission
In addition to this advantage, the tilting mechanism turns the conventional quadcopter,
which is an under-actuated vehicle, to an over-actuated robot which allows full control over
a wide range of state-space. In order to incorporate this new mechanism, which makes the
dynamics of the quadcopter highly nonlinear, this research focuses on developing novel control
mechanisms in order to achieve the desired flight requirements as well as make the vehicle
1.2 Approaches
In order to accomplish the objectives set forth for this study, the new equations of motion for
the tilting quadcopter are first derived mathematically. The derived equations is used to design
1
a linear and nonlinear controller. The numerical simulation of the platform is programmed in
both Simulink and MATLAB for solving the highly nonlinear dynamic equations of motion.
The simulation environment is then used to verify the performance of the developed control
mechanisms. Once the simulation results indicate achievement of desired performance after
running with different flight situations, a 3D model of the new platform is designed using
SolidWorks. The designed 3D model is used to make and 3D print different parts and fabricate
the vehicle. Several attempts are made to improve the platform design. Each attempt focused
on factors such as the weight, the robustness and the functionality of the platform. Several set
of real-world flight tests are carried out to evaluate the capability of the platform in different
flight conditions. The actual flight tests have validated the derived equations for the dynamic
Quadcopters are one of the most popular designs for miniature aerial vehicles (MAVs)
[56]due to their vertical take-off and landing capability, simplicity of construction, maneu-
verability, and ability to negotiate tight spaces making it possible for use in cluttered indoor
areas. Due to these capabilities, quadcopters have recently been considered for a variety of
applications both in military and civilian domains [21, 49, 18, 12]. In particular, quadcopter
MAVs have been explored for applications such as surveillance and exploration of disasters
[11, 33, 67] (such as fire, earthquake, and flood), search and rescue operations [17, 73], moni-
toring of hazmat spills [2], and mobile sensor networks [19, 72].
Blimps [91], fixed-wing planes, single rotor helicopters, bird-like prototypes [28], coaxial
dual rotor helicopters [53], quad-rotors [68, 69, 70, 71] and tilting rotor quadcopters [51, 55,
54, 26, 61, 78] are examples of different configurations and propulsion mechanisms that have
been developed to allow 3D movements in aerial platforms. Each of these has advantages and
drawbacks. This dissertation focuses on quadcopters or quad-rotors which consist of four rotors
in total, with two pairs of counter-rotating, fixed-pitch blades located at the four corners of the
aircraft. This kind of design has two main advantages over the comparable vertical takeoff and
2
landing (VTOL) Unmanned Aerial Vehicles (UAVs) such as single rotor helicopters. Firstly,
quad-rotors do not require complicated mechanical linkage for rotor actuation. Quad-rotors uti-
lize four fixed pitch rotors the variations of whose speeds form the basis of the control. It results
in simplified design and maintenance of the quad-rotors. Secondly, the use of four individual
rotors results in their smaller diameters as compared to the similar main rotor of a helicopter.
The smaller the rotors the less is stored kinetic energy associated with each rotor. This dimin-
ishes the risk posed by the rotors if it comes in contact with any external object. Furthermore,
by securing the rotors inside a frame, the protection of rotors during collisions is achieved. It
allows indoor flights in obstacle-dense environments with lower risk of quad-rotor damage, and
higher operator and surrounding safety. These benefits have resulted in safe test flight by inex-
perienced pilots in indoor environments and recovery time in case of collisions. In particular,
vertical, low speed, and stationary flight are well-known characteristics of a quad-rotors. Struc-
turally, quad-rotors can be made in a small size, with a simple mechanics and control. Though,
as a main drawback, the high energy consumption can be mentioned. However, the trade-off
results are very positive. This configuration can be attractive in particular for surveillance, for
each motor. The front rotor and back rotor pair rotates in a clockwise direction, while the
right rotor and left rotor pair rotates in a counter-clockwise direction. This configuration is
devised in order to balance the moment created by each of the spinning rotor pairs. There are
basically four maneuvers that can be accomplished by changing the speeds of the four rotors.
By changing the relative speed of the right and left rotors, the roll angle of the quad-rotor is
controlled. Similarly, the pitch angle is controlled by varying the relative speeds of the front and
back rotors, and the yaw angle by varying the speeds of clockwise rotating pair and counter-
clockwise rotating pair. Increasing or decreasing the speeds of all four rotors simultaneously
One of the basic limitations of the classical quad-rotor design is that by having only 4
independent control inputs, i.e., the 4 propeller spinning velocities, the independent control of
the six-dimensional position and orientation of the quad-rotor is not possible. For instance, a
3
quad-rotor can hover in place only and if only when being horizontal to the ground plane or it
needs to tilt along the desired direction of motion to be able to move. Tilting rotor quadcopter
Tilt-design makes the dynamics of the quadcopter more complex, and introduces additional
challenges in the control design. However, tilting rotor quadcopter, designed by using addi-
tional four servo motors that allows the rotors to tilt, is an over-actuated system that potentially
can track an arbitrary trajectory over time. It gives the full controllability over the quad-rotor
Another application of the tilting platform lies in its ability to recover during a failure situa-
tion. In conventional quadcopters, if one of the propellers fails, due to its inherent dependency
on the symmetry of the platform, the vehicle becomes entirely uncontrollable. However, in the
proposed platform, if one of the motors completely fail, by using the tilt mechanism of one of
the three remaining motors, the unbalanced momentum can be compensated. Moreover, pro-
viding additional actuation would make the quadcopter more robust to disturbances which can
be rejected more effectively because of the enhanced maneuverability of the quadcopter with
tilting design.
There is a lot of interest recently in developing small aerial vehicles that can carry humans
for transportation in an autonomous manner. To have the quadcopter to be operational for such
purposes, it needs to be scaled up to be able to carry more payloads. Due to safety issues of
the human passenger, it should be robust to external disturbances and unpredictable situations.
To reject external disturbances, agility becomes an important issue. The problem with scaled
up quadcopter is the heavy weight. Once the weight scales up, the inertia will increase and
larger moments will be needed to create the same angular acceleration on a lighter vehicle. As
rotate the new propellers which add more weight to the vehicle. This would make the vehicle
to become more sluggish. Although the tilting mechanism can not make a big difference in
throttle, it is expected that the orientation agility can be increased in big size vehicles due to
4
1.4 Contribution
• The mathematical representation of the quadcopter dynamics with tilting rotors has been
• Appropriate control techniques have been designed for highly nonlinear dynamics of the
• Based on the dynamic equation of motion of tilting rotor quadcopter, the dynamic model
of quadcopter with one motor failure has been derived and the control technique has been
designed in order not only to maintain the stability of the vehicle after the failure, but also
• Two different platforms have been fabricated for the quadcopter which were designed in
SolidWorks environments and some parts have been printed by using 3D printer.
• The numerical simulations and experimental results have validated the mathematical rep-
1.5 Publications
Journals
Dynamic Optimal UAV Trajectory Planning in the National Airspace System via Mixed
5
• Alireza Nemati and Manish Kumar. Dynamic Modeling and Control of a Quadcopter
with Tilting Rotors, submitted to IEEE Transactions Aerospace and Electronic Systems,
2015.
Book Chapters
• Manish Kumar, Mohammad Sarim and Alireza Nemati. Autonomous Navigation and Tar-
get Geolocation in GPS Denied Environment, Multi-Rotor Platform Based UAV Systems.
Wiley Publishing,
• Manish Kumar, Alireza Nemati, Anoop Sathyan and Kelly Cohen. Real-time Video and
FLIR Image Processing for Enhanced Situational Awareness, Multi-Rotor Platform Based
Proceedings
• Alireza Nemati and Manish Kumar. Modeling and Control of a Single Axis Tilting Quad-
• Alireza Nemati and Manish Kumar. Non-Linear Control of Tilting Quadcopter Using
Feedback Linearization Based Motion Control. Dynamic System and Control Conference
Flight Formation of Quad -copters in Presence of Dynamic Obstacles using Mixed Integer
6.2015-0715, 2015.
6
• Mohammad Sarim, Alireza Nemati, Manish Kumar and Kelly Cohen. Extended Kalman
Filter based Quadrotor State Estimation based on Asynchronous Multisensor Data, Dy-
• Mohammadreza Radmanesh, Manish Kumar, Alireza Nemati and Mohammad Sarim. So-
lution of Traveling Salesman Problem with Hotel Selection in the Framework of MILP-
• Alireza Nemati, Neal Soni, Mohammad Sarim, and Manish Kumar. Design, fabrication
and control of a tilt rotor quadcopter. In ASME 2016 Dynamic systems and control con-
• Alireza Nemati, Rumit Kumar, and Manish Kumar. Stabilizing and control of tilting-rotor
quadcopter in case of a propeller failure. In ASME 2016 Dynamic systems and control
Intellectual property
• Alireza Nemati, Mehdi Hashemi and Manish Kumar,“ World Frame Based Radio Con-
troller(RC) for Multi-copter UAVs” Filed for provisional patent, October 2015
• Alireza Nemati, Manish Kumar and Rumit Kumar. Fault Tolerance Quadcopter . Provi-
This dissertation consists of eight chapters. This includes “Introduction” as the first chapter.
Chapter 2 is a literature review that provides a brief history of the conventional as well as tilting
quadcopters. Chapter 3 reports on the dynamic model of the tilting quadcopter and considers
the nonlinearities that add to the equation due to additional control inputs. Chapter 4 presents a
combined linear and nonlinear controller which is used to control desired orientation during the
flight. The dynamic model of the vehicle and the proposed control technique once the failure
7
The hardware design process is described in detail in Chapter 6. Results from numerical
simulation and experimental studies carried out to verify the modeling and control of tilting
rotor quatcopters following a reference trajectory with simultaneous control of both pitch and
roll angles are discussed in Chapter 7. The conclusions and future works are presented in Chap-
ter 8. This chapter summarizes the dissertation, discusses the contributions and also outlines
8
Chapter 2
and Landing (VTOL) that is lifted or propelled by four independently rotating rotors. The
idea behind Quadcopters was first developed in the early 1900s. There were very few unique
and momentous quadcopter designs developed throughout the 20th century. The earliest ideas
for a quadcopter were designed and test piloted by Louis Breguet, Etienne Oemichen, George
DeBothezat, and D.H. Kaplan. The first successful flight of a quadcopter aerial vehicle was in
1907. This device, named the Gyroplane (Figure (2-1) 1 ), was built by Breguet brothers and
consisted of a 55hp Renault engine and two forward-tilting 2-blade rotors. It was reported to
have multiple successful flights during the summer of 1908. However it’s mobility and range
1
https://en.wikipedia.org/wiki/Breguet-Richet-Gyroplane
9
Figure 2-1: Gyroplane.
In the 1920s, Etienne Oemichen, was able to construct the first stable VTOL quadcopter
which he named Oemichen II (Figure (2-2) 2 ). It consisted of a single 180hp Gnome engine
powering four rotors, a complex steel-tube framework of cruciform layout, five smaller pro-
pellers mounted horizontally to provide lateral stability, and an additional pair of propellers
that were mounted to the nose of the craft for steering. The last pair of propellers provided
forward thrust. This design made thousands of successful flights during the mid 1920s and
even established a world record of flying one kilometer in seven minutes and forty seconds.
Almost all quadcopters in the 1920s were unable to sustain a controlled flight and had to use
the Ground Effect to sustain flight limiting these designs to stay low to the ground.
Around the same time, George DeBothezat designed and built the Flying Octopus (Figure
3
(2-3) for the United States Army Air Corps. The 1678kg ”X” shaped structure supported
2
https://en.wikipedia.org/wiki/C389tienne-Oehmichen
3
https://en.wikipedia.org/wiki/Pescara-Model-3-Helicopter
10
four 8.1m diameter six-blade rotors; one on each end of the 9m long arms. At each end of the
lateral arms, two smaller propellers with variable pitch supplied thrust and enabled yaw control.
After working on his design for a little over two years, DeBothezat was able to develop a fairly
capable quadcopter. This design was able to carry a payload of up to 4 people including the
pilot. However, the design was considered to be flawed as it was under-powered, unresponsive
and very fragile. The craft was only capable of reaching an altitude of around 5m rather than
Early quadcopters typically contained a single engine positioned in the center of the fuse-
lage that drove the four rotors via belts or shafts. These belts and shafts were heavy and more
importantly broke down often. In addition, the four rotors of the quadcopter were ever so
slightly different from one another, so the quadcopter was not naturally stable during flight.
Running all rotors at the same speed did not produce a stable flight and each rotor had to be
constantly adjusted to sustain a stable flight. In the early 1900s, with the absence of any dig-
ital computers or sensors, flying a quadcopter required a monumental workload for the pilot
making the early quadcopters very inefficient and not practical for transportation. These early
quadcopters designs also included multiple additional rotors located on different locations of
the quadcopter for additional stability, making these designs not true quadcopters. As mate-
rials and engineering practices evolved over the century, numerous improvements were made
by both increasing the power of the motors and reducing the overall weight of the designs.
During the early 1950s, D.H. Kaplan, worked on and test piloted the Convertawings Model A
11
Quadcopter (Figure (2-4) 4 ). Kaplan’s design featured four rotors and had a two motor layout
with the rotors positioned in an H configuration. Kaplan’s machine may be considered the first
true quadcopter as it was capable of sustaining a controlled flight without the use of the ground
effect or any additional propellers. The 2, 200 pounds craft had a much simpler design than pre-
vious quadcopters due to the fact that control was obtained by varying the thrust between the
individual rotors eliminating the need for complex cyclic-pitch-control systems and additional
rotors on the sides of the fuselage. This design first flew in March of 1956 with great success.
The design, in particular its control system, was a precursor of a majority of the current vertical
Tilt Rotors combined the properties of a helicopter which included Vertical Take Off and
Landing (VTOL), hovering, and vertical, forward, and lateral flight, with the desirable prop-
erties of a fixed-wing aircraft including long range flight, low power consumption and the
ability to carry heavier payloads. The first design that resembled a modern tilt rotor device was
4
https://en.wikipedia.org/wiki/De-Bothezat-helicopter
5
http://history.nasa.gov/monograph17.pdf
12
Figure 2-5: George Lehbergers 1930 tilting propeller vertical take-off “flying machine”.
Though this design never amounted to a prototype, it was the first step in making a func-
tional VTOL capable tiltrotor vehicle and inspired the design of the Focke-Achgelis FA-269
(Figure (2-6)6 ) trail-rotor convertiplane project in Germany during World War II[43].
6
http://history.nasa.gov/monograph17.pdf
13
Figure 2-6: Three-view drawing of the Focke-Achgelis FA-269 convertiplane
A prototype of this aircraft was built in 1943 and consisted of two pusher propellers that
tilted below the two wings for takeoff and landing. However, the project was discontinued
after the allies destroyed a full scale mock-up of this design and much of the research during
a bombing in 1944. A few years later, variants of this tilt rotor configuration surfaced again in
The Bell XV-3 (Figure (2-7)7 ) was a tiltrotor aircraft designed by Bell in the 1950s[15].
Its first successful flight was in August 1955. It was the first aircraft to successfully transition
between helicopter and fixed wing for normal flight. The XV-3 was powered by a single 450hp
radial engine that propelled the aircraft at a maximum speed of 296 km/h. The craft had a
maximum altitude of 4600 meters. This aircraft was a proof of concept and made over 100
successful transitions before it was severely damaged in a wind tunnel accident and the design
was scraped.
7
http://ntrs.nasa.gov/archive/nasa/casi.ntrs.nasa.gov/20040087005.pdf
14
Figure 2-7: The Bell XV-3, during flight testing.
The data and experience collected during this trial were key to the development of the Bell-
XV15 (Figure (2-8) 8 ) and the V-22 Osprey (Figure (2-9) 9 )[57, 43].
8
https://en.wikipedia.org/wiki/Bell-XV-3
9
https://en.wikipedia.org/wiki/Bell-Boeing-V-22-Osprey
15
Figure 2-9: The V-22 Osprey, during transition flight
Both these designs followed the same principles as the Bell XV-3 and had many successful
In respect to quadcopter tilting machines, there have been two early designs that stand out,
the Curtiss X-19 and the and the Bell X-22. The Curtiss X-19 (Figure (2-10)10 ) built in 1960s
was a passenger plane that consisted of two sets of thin wings each with a 3 bladed rotor that
could rotate 90 degrees [30]. With its massive 2,200 hp engines, it could carry up to 550kg of
cargo or 4 passengers along with the two crew members. Two turboshaft engines were housed
in the rear fuselage and powered the four rotors. The aircrafts first flight was in 1963 and was
capable of flying up to 523 Km and reached a maximum speed of 650 Km/h. Two prototypes
of the X-19 were built but the project was canceled after the first prototype crashed during its
second flight.
10
https://en.wikipedia.org/wiki/Curtiss-Wright-X-19
16
Figure 2-10: X-19 in hovering flight
The Bell X-22 was built a couple years after the Curtiss X-19 and is considered to be one of
the most versatile and longest lived of the many VTOL aircrafts that have been developed[30].
It is similar to the X-19 in that it has four wings each with their own 3 bladed propeller each able
to rotate 90 degrees but instead of having 2 motors, the Bell X-22 had four 1250 hp motors each
powering their own rotor. The design was able to carry up to six passengers and two pilots and
reached a maximum speed of 507 km/hour with a range of up to 716km. The two prototypes
of the Bell X-22 were used for many years by both NASA and the US Navy for V/STOL and
performed very well. One is still on display at the Niagara Falls Aerospace Museum in New
York (Figure (2-11)11 ). More modules of quadcopters are currently being developed for the US
Army Corps Including the Bell Boeing Quad Tiltrotor (QTR). It is currently under study and
was first designed in 1999. The Bell Boeing Quad Titlrotor is predicted to be able to carry up
11
https://en.wikipedia.org/wiki/Bell-X-22
17
2.2 Current Quadcopters
Quadcopters are one of the most popular designs for miniature aerial vehicles (MAVs) due
to their vertical take-off and landing capability, simplicity of construction, maneuverability, and
ability to negotiate tight spaces making it possible for their use in cluttered indoor areas. Due to
these capabilities, quadcopters have recently been considered for a variety of applications both
in military and civilian domains. In particular, quadcopter MAVs have been explored for appli-
cations such as surveillance and exploration of disasters (such as fire, earthquake, and flood),
search and rescue operations, monitoring of hazmat spills, and mobile sensor networks[37]
[64][14]. Blimps, fixed-wing planes, single rotor helicopters, bird-like prototypes, coaxial dual
rotor helicopters, quad-rotors, tilting rotor quadcopters are examples of different configurations
and propulsion mechanisms that have been developed to allow 3D movements in aerial plat-
forms [39] [5] [90] [50] . Each of these designs have advantages as well as drawbacks. This
work focuses on quadcopters or quad-rotors which consist of four rotors in total, with two pairs
of counter-rotating, fixed-pitch blades located at the four corners of the aircraft. This kind of
design has two main advantages over the comparable vertical takeoff and landing (VTOL) Un-
manned Aerial Vehicles (UAVs) such as single rotor helicopters. Firstly, quad-rotors do not
require complicated mechanical linkage for rotor actuation. Quad-rotors utilize four fixed pitch
rotors the variations of whose speeds form the basis of the control. It results in simplified design
and maintenance of the quad-rotors. Secondly, the use of four individual rotors results in their
smaller diameters as compared to the similar single rotor of a helicopter. Smaller rotors imply
less stored kinetic energy associated with each rotor. This diminishes the risk posed by the ro-
tors if it comes in contact with any external object. Furthermore, by securing the rotors inside
a frame, the protection of rotors during collisions is achieved. This configuration allows indoor
flights in obstacle-dense environments with lower risk of quad-rotor damage, and higher oper-
ator and surrounding safety. These benefits have resulted in safe test flights by inexperienced
pilots in indoor environments and lesser recovery time in case of collisions [25]. In particular,
stable, vertical, low speed and stationary flights are well-known characteristics of a quad-rotor.
Structurally, quad-rotors can be designed in a small size, with simple mechanics and control.
The quadcopters have been found to be an attractive choice in particular for surveillance, for
18
imaging dangerous environments, and for outdoor navigation and mapping [59] , [22]. The
major drawback, however, is high energy consumption due to the use of four rotors.
each motor. The front rotor and back rotor pair rotates in a clockwise direction, while the
right rotor and left rotor pair rotates in a counter-clockwise direction. This configuration is
devised in order to balance the moment created by each of the spinning rotor pairs. There are
basically four maneuvers that can be accomplished by changing the speeds of the four rotors.
By changing the relative speed of the right and left rotors, the roll angle of the quad-rotor is
controlled. Similarly, the pitch angle is controlled by varying the relative speeds of the front and
back rotors, and the yaw angle by varying the speeds of clockwise rotating pair and counter-
clockwise rotating pair. Increasing or decreasing the speeds of all four rotors simultaneously
One of the basic limitations of the classical quad-rotor design is that by having only 4
independent control inputs, i.e., the 4 propeller spinning velocities, the independent control of
the six-dimensional position and orientation of the quad-rotor is not possible. For instance, a
quad-rotor can hover in place only and if only while being horizontal to the ground plane or it
needs to tilt along the desired direction of motion to be able to move. Tilting rotor quadcopter
design has been developed to solve these basic limitations of a quad-rotor. In next section, there
Recently, there has been a renewed interest in quadcopters from hobbyists, universities, and
corporations across the globe. The renewed interest is due to the many significant technological
advances in sensors and micro-controllers over the past decade that have allowed these once
large machines to be miniaturized to fit in the palm of a hand and be autonomously controlled.
A significant number of quadcopters have been introduced for both military and civilian use as
a result of partnerships between companies and universities that have enabled this quadcopter
UAV revolution. Many companies such as AeroQuad, ArduCopter, DJI, and Parrot AR.Drone
have sparked the interest of hobbyist; Coupled with the DIY (Do It Yourself) and open source
movement, Quadcopter UAVs are more popular and progress in the sector is advancing faster
19
2.2.1 Quadcopters with Tilting Mechanism
Quadcopters with tilting propellers are divided into two categories. Quadcopters which
have the capability of flying both as a conventional quadcopter as well as a fixed wing aerial
vehicle by tilting all propellers in the same direction by the same amount of angle and the
one which is capable of making a tilt in any individual propellers independently. In the first
type of tilting quadcopter( Convertible Quad Tilt Rotor (CQTR)), all propellers are changing
their orientation simultaneously with the same amount of angle. However, variety of platform,
mechanism and control methods have been used by different researchers and universities. Since
all propellers tilt with the same angle, the dynamic equation of the aircraft does not change a lot
and the complexity of the equation is not the most critical obstacle that needs to be tackled for
the designers. The ability of transition between vertical take off and hover flight to horizontal
flight is one the most difficulties the needs to be consider for CQTR. With this ability, aircraft
will be able not only to take off and land at any inappropriate area, but also by taking an
advantage of aerodynamic shape of the wing, it will be able of flying horizontally in long
distances. The flight mode that makes the transition between vertical and horizontal flight has
been gaining remarkable interest among researchers. Numerous innovative CQTR platform
have been studied in very last few years. Papachristos et al [62] from university of Patras,
Greece, have focused on hybrid model predictive flight mode conversion control of CQTR.
capable of flying both as a quadcopter as well as fix wing aircraft. They have developed
12
http://www.nt.ntnu.no/users/skoge/prost/proceedings/ecc-2013/data/papers/1271.pdf
20
nonlinear dynamics has been derived and used as a model for control during autonomous mid-
flight conversion. Although they have not flown the aircraft, but by using simulation studies its
been shown that their proposed strategy exceeds the functionality of the flight-modes conver-
sion. The standard NACA2411 airfoil has been selected for the design. The wings with total
span of 1 meter are mounted on the tilting mechanism. The wings are capable of rotating 90
degrees angle.
Another research group has focused on design and control of gas-electric hybrid CQTR
with morphing wing in order to extend the hovering flight up to 3 hours or up to 10 hours of
horizontal flight [13]. They have minimized the mechanical morphing wings and aerodynamic
cost for both high speed and low speed flights. A variety of novel features have been used in
A gas engine-battery hybrid propulsion is used due to capability of carrying heavy payload
and very long flight duration, plus a carbon composite wings not only to reduce the weight
but also for handling both high speed and low speed flight. The V-type structure of the aircraft
allows to extension of the rear rotors from the center of the mass which ensure the full coverage
of the wing area by the rotor for preventing stalls as well as minimum required speed for
horizontal flight. For the transition, two servo motors have been located next to shaft of rear
and front rotors. The vehicle’s take off weight is 20 kg and the wing-span including the larger
winglets is 2.5 meters. Numerical simulation has been used to validate and understand the
flight behavior and performance of the aircraft. By placing four rotors in two axes with almost
the same level, rotor’s thrust will not cover the most, even if like the previous work, the rotors
13
http://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=7152278
21
of one the axis extend from the center. The researchers in Beihang University, China, [65] have
proposed a prototype concept of CQTR that place the rotors of the front and rear axis of the
horizontal mode in two different levels. In their concept, for change in configuration, the two
front rotors tilt down to -90 degree, while the two rear rotor tilt up to 90 degree. After the
transition is done, all rotors will be facing front with exactly the same angle but in two different
14
levels. Figure (2-14) shows the configuration of the CQTR in both quadcopter and flying
wing mode.
Figure 2-14: The Configuration of the QTR UAV under research a) Quadcopter mode b)Flying
They have focused on the trajectory tracking control for hovering and acceleration flight of
CQTR by using dynamic inversion. Their scenario is based on 4 strategies as follows: take-
off and reach a certain altitude, find an optimal transition trajectory to not only minimize the
transition time but also not to lose altitude, the next scenario is to keep flying in fixed wing mode
and the ability to change the altitude. During the cruise flight, the rotor speed and the forces
allocated between the rotors and the wings maintain the needed thrust to control altitude and
attitude. Attitude Command Attitude Hold (ACAH) method is being used for attitude control
strategy. The last scenario is to lower the speed and keep tilting back the rotors to their original
angle during the takeoff. Although they have built a prototype version of the vehicle with
dimension of 1.8 meter for the wing span and gross weight of 5.2kg, but numerical simulation
Several researches have introduced many interesting prototype of CQTR, but there are few
14
http://comb.buaa.edu.cn/PUBLICATIONS/PATEERS/2014/48.html
22
groups that validate their control systems by experimental results. Hancer et al [23] presented
a prototype CQTR equipped with robust position controller to track desired trajectory under
Dryden model has been used to model wind effects which is included in dynamic model.
These disturbances are being estimated by using disturbance observer which is commonly used
in motion control systems. Parametric uncertainties and nonlinear term are also added to ex-
ternal disturbances as a total disturbance. Performance of the hovering flight is verified with
the experimental results. The transition from hovering flight to horizontal cruise flight which
is the most critical part of the experiment is not tested in real world, but trajectory tracking
Mikami and Uchiyama [47] from Nihon University have validated their concept by nu-
behavior, a linearization method without any approximation has been applied to their control
strategy. This proposed control strategy is being used in both translational and rotational con-
15
http://research.sabanciuniv.edu/15316/1/cdc10.pdf
16
http://people.sabanciuniv.edu/munel/Publications/JournalPapers/Mechatronics-2012.pdf
17
http://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=7152364
23
Figure 2-16: Overview of developed CQTR, Nihon University.
Overall weight of the prototype is 0.48 kg with the length and wing span of 0.8 and 0.7
meter, respectively. Although, due to weight limitation and small size of the battery, the en-
durance of the vehicle is not longer than 15 minutes, but the experimental results verified the
validity of flight control strategy. The vehicle is equipped with micro-computer, flow sensor,
radio module, IMU and ultra-sonic sensor. The experimental results show that the transition
Even-though above mentioned air-crafts are using a tilting mechanism for the propellers,
but the under actuation problem of the robot still remains, and in addition to that, these ad-
ditional dynamic of convertible quad tilt rotor is not making the big problem for the control
systems. Also by changing the angle of all propellers simultaneously, translational motion can
not be achieved independently without tilting the aircraft. If the aircraft’s arm had the capabil-
ity of independently tilting about their own axis, the dependency of translation and orientation
problem, could be solved and the system would not be called an under actuated robot. There
have been some attempts to carryout this problem by several groups. Ryll et al. [76] from
Max Plank Institute, Germany, proposed a tilting quadcopter with 8 independent control inputs
that allows the aircraft for independent attitude and position control. They have added four
additional servo motors to allow the propellers to tilt about their axes18 .
18
https://www.semanticscholar.org/paper/Modeling-and-control-of-a-quadrotor-UAV-with-Ryll-
BC3BClthoff/3508f3f9497425a4cbc4bc2d2c2d30df3fc6be77
24
Figure 2-17: CAD model of the quadcopter with tilting propellers, Max Planck Institute.
The linearized compensation control based on the quadcopter’s dynamic has been used
to achieve 6 DOF of motion control. The tracking of an arbitrary trajectory was the main
focus of the group. They have simplified the dynamic model in order to have a suited model
for control design. Since 8 independent inputs are available, their proposed control design is
over actuated. As in many output tracking control techniques, an appropriate way they solved
the problem was to place the output feedback linearization method. Numerical simulation
was the way they applied the controller presented in [76]. The tracking performance of the
controller was validated and the capability of the proposed method to avoid the singularities
was guaranteed. Another paper has been published from the same authors [77] to validate the
proposed strategies. Although the prototype which is called ”omnicopter” has been tested on a
testing gimbals, but their results show the full controllability over 6 DOF body pose in space.
Another group [81] has tested their prototype on the ball joint rig. Ball joint rig is a device
that can be attached to the underneath of the quadcopter and it can let the quadcopter to orient
19
https://www.semanticscholar.org/paper/A-Novel-Actuation-Concept-for-a-Multi-Rotor-UAV-Segui-Gasco-
Al-Rihani/28bdc9144a153c132b37382be32e94d83b1abfed/pdf
25
Figure 2-18: Vehicle prototype on the ball joint rig flight test, Cranfield University.
The main proposal of the prototype was to improve the performance and fault tolerance of
quadcopter vehicles. They have proposed dual axis tilting propellers which enables gyroscope
torque, differential thrusting and thrust vectoring. Not only the mathematical representation of
the model was modeled and verified by the experimental results, but also a control system was
developed based on PD controller and validated through test on a ball joint rig flight test.
Tilt-design makes the dynamics of the quadcopter more complex, and introduces additional
challenges in the control design. However, tilting rotor quadcopter, designed by using addi-
tional four servo motors that allow the rotors to tilt, is an over-actuated system that potentially
can track an arbitrary trajectory over time. It gives the full controllability over the quad-rotor
position and orientation providing possibility of hovering in a tilted configuration. This work
presents a mathematical dynamic modeling of the tilting rotor quadcopter which provides a
description of the dynamical behavior of the quadcopter as a function of the rotational speeds
of each of the four rotors and their respective tilt-angles. The developed mathematical repre-
sentation of the tilting rotor quadcopter can be used to obtain the position and orientation of
the quad-rotor. The same model can further be used to develop a linear and nonlinear control
strategy via which the speeds of the individual motors and the respective tilt-angles can be ma-
26
Chapter 3
Dynamic Modeling
Unlike traditional quad-rotor models, which have only four rotatory propellers as the vehi-
cle’s inputs, in tilting rotor quadcopters, there are four more servo motors attached to the each
arm that adds one degree of freedom to each of the propellers, resulting in the tilting motion
along their axes. In this chapter, first the dynamic model of a traditional quad-rotor is described,
Figure (3-1) schematically shows the coordinate system and forces acting on a traditional
quad-rotor. In the 3 dimensional space, the world-frame (E) denotes the fixed reference frame
with respect to which all motions can be referred to and the body-frame (B) is a frame attached
to the center of mass of the vehicle. The rotation of each rotor causes an aerodynamic force
or thrust that acts perpendicular to the plane of rotation of the rotor. In addition to the forces,
each rotor produces a moment perpendicular to the plane of propeller rotation. The moment
produced by a propeller on the vehicle is directed opposite to the direction of rotation of the
propeller, and therefore to cancel out rotation along the Z-axis, the moments for rotor 1 and
3 are set in clockwise (−ZB ) direction and for rotor 2 and 4 are set in counter clockwise (ZB )
direction.
Based on NASA Standard Airplane [35], Euler angle transformations are defined by ψ,
θ and φ which respectively represents the heading, attitude and bank angles also referred to
as yaw, pitch and roll angles. Combined transformation matrix from body coordinate to the
27
world coordinate is obtained by three successive rotations. The first rotation is about X axis,
followed by another rotation about Y axis and the last rotation is about Z axis. For each rotation,
cψcθ cψsθsφ − sψcφ cψsθcφ + sψsφ
REB = sψcθ sψsθsφ + cψcφ sψsθcφ − cψsφ (3.1)
−sθ cθsφ cθcφ
where cψ and sψ denote cos(ψ) and sin(ψ) respectively, and similarly for other angles.
By obtaining vehicle’s vertical forces in the world frame and writing the equations of mo-
tion based on the Newton second law along the X, Y and Z axes, we can write:
X
m ẍ = Fi (sψsφ + cψsθcφ) − C1 ẋ
X
mÿ = Fi (sψsθcφ − cψsφ) − C2 ẏ (3.2)
X
mz̈ = Fi (cθcφ) − mg − C3 ż
where m is the total mass of quad-rotor, g is the acceleration due to gravity, x, y and z are
quadcopter position in world frame coordinate, C1 , C2 and C3 are drag coefficients. Note that
28
the drag forces are negligible at the low speed. Fi ,(i = 1, 2, 3, 4) are forces produced by the
Fi = K f ω2i (3.3)
where ωi is the angular velocity of ith rotor and K f is a constant. In addition, Euler equations
are written in order to obtain angular accelerations of the vehicle given by:
Iz ψ̈ = M1 − M2 + M3 − M4 − C30 ψ̇
where l is distance of each rotor from the vehicle’s center of mass. I x , Iy and Iz are moment of
inertia along x, y and z directions respectively. C10 , C20 and C30 are rotational drag coefficients.
Mi , (i = 1, 2, 3, 4) are rotors moment produced by angular velocity of rotors and given by:
Mi = Km ω2i (3.5)
Figure 3-1: Schematic diagram showing the coordinate systems and forces acting on the quad-
rotor
During a hovering flight, the quad-rotor not only has zero acceleration and velocity but also
29
needs to have zero pitch and roll angles, i.e. r = r0 , θ = φ = 0, ψ = ψ0 , ṙ = 0, θ̇ = φ̇ = ψ̇ = 0.
At this nominal hover state, the produced force form each propellers must satisfy:
1
Fi = (mg) (3.6)
4
r
mg
ωi = ωh = (3.7)
4k f
For a tilting rotor quadcopter, four other variables are added representing the angles of the
quad-rotor arms. Adjustment of these angles results into improved vehicle maneuverability and
To illustrate the motion of the tilting rotors quadcopter, a schematic diagram showing the
forces/moments acting and coordinate frames used in the modeling is provided in Figure (3-2).
As it can be seen from this figure, the propellers are free to tilt along their axes. The planes
shown with dashed lines are the original planes of rotation with zero tilt angles for the respective
propellers. Similarly, the planes shown with the rigid lines are the tilted planes of rotation for
may be noted that the forces generated by the propellers are perpendicular to these respective
planes of rotations.
30
Figure 3-2: Coordinate Frames and Free body diagram of Tilting Quadcopter
ẍ 0 F1 sθ1 − F3 sθ3 C1 ẋ
m ÿ = 0 − REB − C2 ẏ
F 2 sθ1 − F 4 sθ3
F1 cθ1 + F2 cθ2 + F3 cθ3 + F4 cθ4 C3 ż
z̈ −mg
Using the rotational matrix in (1), equations of motion in world-frame can be rewritten as:
31
mz̈ = −F1 sθ1 sθ + F3 sθ3 sθ − F4 sθ4 cθsφ
where M 0i , (i = 1, 2, 3, 4) are the tilting moments created by the four servo motors attached
to the end of each arm to enable their tilting motion. It may be noted that these moments are
negligible because the moments produced by the servo motors are used to tilt the arms which are
connected to the main body via mechanical bearings. Neglecting bearing friction, the moments
transmitted to the main body of the quadrotor are negligible. Based on the dynamic model
presented above, we propose the following two Theorems. Note that without lost of generality,
Theorem 1: Considering the dynamics of the tilting rotor quadcopter given by Equations
(3.8) and (3.9), and assuming the relationship between the tilting angles of the four rotors
θ1 = −θ3 and θ2 = −θ4 and all rotors having equal rotational speeds, the quadcopter, at an
equilibrium hovering state, achieves a roll angle φ given by φ = θ1 /2 when the pitch angle is
zero, and a pitch angle θ given by θ = θ2 /2 when the roll angle is zero.
Proof: In tilt-hovering, the arm angles of the first and third propellers are tilted by θ1 and
θ3 = −θ1 , respectively. This produces a roll angle φ of the vehicle, and, the equations for linear
32
motion of the quadcopter is given by:
F1 s(θ1 − φ) + F3 s(−θ3 − φ) − F2 sφ − F4 sφ 0
m ẍ
mÿ = 0 +
0
(3.10)
mz̈ F c(θ − φ) + F cφ + F c(−θ − φ) + F cφ −mg
1 1 2 3 3 4
For hovering, the accelerations ẍ, ÿ, and z̈ should all be equal to zero. Using the equation
rotational speeds of all rotors are the same, the angle φ can be obtained as:
θ1
φ= (3.11)
2
Similarly to the equation (3.10), if the second and fourth arms are tilted, the equations of motion
can be written as:
m ẍ 0 0
mÿ = −F sθ − F sθ + F s(θ − θ) + F s(−θ − θ) + 0 (3.12)
1 3 3 2 2 4 4
mz̈ F cθ + F c(θ − θ) + F cθ + F c(−θ − θ) −mg
1 2 2 3 4 4
Similar to above, the angle θ resulted from tilting of the second and forth arms, is given by:
θ2
θ= (3.13)
2
Theorem 2: Considering the dynamics of the tilting rotor quadcopter given by Equations
(3.8) and (3.9), and assuming the relationship between the tilting angles of the four rotors
θ1 = −θ3 and θ2 = −θ4 , the motor speed needed for vehicle for hovering with a tilt angle is
given by:
v
t mg
u
ωi = ωh = when θ = 0
θ1
4k f c
2
and
v
t mg
u
ωi = ωh = when φ = 0 (3.14)
θ2
4k f c
2
33
Proof: In hovering with roll angle and zero pitch angle, the acceleration along z axis is zero,
θ1 X
cos( ) Fi = mg (3.15)
2
Based on Equation (3.3), and noting that each rotor’s angular speed is the same (i.e., F1 = F2 =
v
t mg
u
ωi = ωh = (3.16)
θ1
4k f c
2
Similarly, considering hovering with pitch angle and zero roll angle, the Equation (3.12) gives:
θ2 X
cos( ) Fi = mg (3.17)
2
Now similar to above, the angular speeds of the rotors is given by:
v
t mg
u
ωi = ωh = (3.18)
θ2
4k f c
2
34
Chapter 4
Control System
In this chapter, the control strategy of the tilting rotor quadcopter is presented. The aim of
the control strategy is not only control the position of the vehicle to follow an arbitrary trajec-
tory in 3 dimensions, but also to have control over the orientation of the vehicle in hovering as
The controller inputs are four independent speeds of propellers and their rotations about the
axes of quadcopter arms. Referring to Figure (3-2) and the two Theorems, it is assumed that
θ1 = −θ3 and θ2 = −θ4 . It may be noted that these constraints, in fact, make the over-actuated
system into fully actuated system (two inputs to tilt the rotors another four inputs for their
rotational speeds make the total number of independent control inputs to be six). For 6 DoF
quadcopter, this results into complete control over its position and orientation. The dynamic
model of the tilting rotor quadcopter, described in (3.8) and (3.9), is used to design the PD
controllers for orientation adjustment and trajectory tracking. Figure (4-1) showes the block
diagram of the control algorithm for orientation and position control during the flight.
35
Figure 4-1: The block diagram of position and orientation control algorithm
To start tracking a specific trajectory, first, hovering from the initial starting point is neces-
sary. Then, the orientation of the vehicle to a specific pitch or roll angle is obtained. In [46],
the relationship between the rotational speeds of the motors and the deviation of the orienta-
tions from nominal vectors for hovering and navigation is described in detail for conventional
quadcopter. Similar to that approach for the tilting rotor quadcopter, the rotational speeds are
observed as:
ω1 1 0 −1 1 ωh + ∆ω f
des
ωdes 1 1 0 −1 ∆ω
2 φ
= (4.1)
ωdes 1 0 1 1 ∆ωθ
3
ωdes 1 −1 0 −1 ∆ωψ
4
where ωdes
i , (i = 1, 2, 3, 4) are the desired angular velocities of the respective rotors. The
hovering speed, ωh , is calculated from Theorem II. The proportional- derivative laws are used
to control ∆ωφ , ∆ωθ , ∆ωψ and ∆ω f which are deviations that result into forces/moments causing
36
roll, pitch, yaw, and a net force along the zB axis, respectively, which are calculated as:
where p, q and t are the component of angular velocities of the vehicle in the body frame. The
relationship between these components and derivatives of the roll, pitch and yaw angles are
φ̇
p
θ̇ = T q
ψ̇
t
1 tanθ.sinφ tanθ.cosφ
T = 0
cosφ −sinφ
0 secθ.sinφ secθ.cosφ
The relationship between the tilt angles of individual rotors, given by θides , i = 1, 2..4, and
θ1 1
des des
0 1 0 2φh
des
θ2 0
des
1 0 1 2θh
= (4.3)
θ3 1 0 ∆φh
des
0 1
θdes 0 1 0 1 ∆θ
4 h
where φdes
h and θh are reference roll and pitch angles and ∆φh and ∆θh are orientation devia-
des
tions. Figure (4-2) shows the orientation of the vehicle with respect to the tilted propellers. A
proportional-derivative controller is used to control the orientation deviation using the reference
37
orientation values as:
des
h − φ) + kd, φh (φh
∆φh = k p, φh (φdes ˙ − p)
des
∆θh = k p, θh (θhdes − θ) + kd, θh (θ˙h − q) (4.4)
The mathematical model of DC servo motor is obtained by the following first order transfer
function that relates the motor angular velocity (rad/s) to input voltage (V) as:
Ω(s) K
= (4.5)
V(s) τs + 1
where τ represents the time constant of the system, and K represents the steady state gain
value. The angular position of the servo motor can be obtained by integrating the motor angular
velocity. The transfer function relating the angular position (rad) and input voltage (V) can be
obtained as:
Θ(s) K K
= = 2 (4.6)
V(s) s(τs + 1) τs + s
The above equation represents a second order transfer function. So, this system is identical
to a second order actuation system. Such systems exhibit a transient response when they are
subjected to external inputs or environmental disturbances. It should be noted that, the transient
response characteristics are one of the most important factors in system design. In general,
38
transfer function of a 2nd order system with input, u(t) and output, y(t) can be expressed as:
y(s) kω2
= 2 (4.7)
u(s) s + 2ζω + ω2
The TGY-210DMH servo motor used in tilt rotor quadrotor system is similar to an actuator
system with damping ratio (ζ) = 0.7, and it has an angular speed of 8 rad/s when operated at
6V and 6.98 rad/s when operating at 4.8V . The natural frequency for mathematical model is
considered to be as 16 rad/s by considering a factor of safety equal to 2, the DC gain has been
considered as unity.
In order to have the quad-rotor track a desired trajectory ri,T , the command acceleration,
where ri and ri,T (i = 1, 2, 3) are the 3-dimensional position of the quad-rotor and desired
During the flight of a tilting quadcopter, the orientation of the vehicle needs to be set at
specific pitch or roll. This can be obtained by linearizing the equation of motion that correspond
or roll angles. The change of the pitch or roll angles are supposed to be small during flight. By
linearizing Equation (3.8) about these nominal hovering states, desired pitch and roll angles to
39
where
A = −s(2φdes
h )c(ψT )s(θh ) + s(2θh )c(ψT )s(φh )cθh
des des des des
+ c(φdes
h )c(ψT )cθh cφh + c(2θh )c(ψT )c(θh )c(φh )
des des des des des
C = s(2φdes
h )c(ψT )c(θh ) + s(2θh )c(ψT )s(θh )s(φh )
des des des des
D = −(2φdes
h )c(ψT )s(θh ) + s(2θh )s(ψT )
des des
F = s(φdes
h )c(ψT )c(θh ) + s(2θh )s(ψT )s(θh )s(φh )
des des des des
φdes and e
where e θdes are respectively the desired deviation in roll and pitch angles from the
40
the orientation is set to be given by the nominal hovering values. Equation (5.14) represents a
φdes and e
pair of two coupled linear equations which are to be solved to obtain the e θdes . The final
φdes = e
φdes + φdes
h
θdes = e
θdes + θhdes (4.10)
The desired speeds of the individual rotors are calculated by Equation (5.12). Equation
The four rotational velocities of the rotors are the inputs of the vehicle, but in order to
simplify the equations of motion which are described in (3.8) and (3.9), new artificial input
variables are defined as the following. It may be noted that we assume that the tilting happens
u1 = (F1 + F2 + F3 + F4 )/m
u2 = l(F3 − F1 )/I x
u3 = l(F4 − F2 )/Iy
The equations of motion of the vehicle, considering small angle assumption and tilting
41
along only roll direction (hence, θ1 = −θ3 and θ2 = θ4 = 0), can be obtained as:
1
ẍ = sinθ1 cosθ + u1 cosθ1 cosφh cosφsinθ
2
ÿ = −u1 cosθ1 sinφh cosφ − u1 cosθ1 cosφh sinφ
1
z̈ = −mg − u1 sinθ1 sinθ + u1 cosθ1 cosφh cosφcosθ
2
− u1 cosθ1 sinφh sinφcosθ
θ̈ = u3
ψ̈ = ku4 (4.12)
where φh is the the desired roll angle that the quadcopter is supposed to tilt during the flight.
A brief review of nonlinear control using feedback linearization method [86, 94, 80, 79]
is presented here. Among the two fundamental design techniques for feedback linearization,
i.e., Input-Output linearization and Input-State linearization, we utilize the Input-Output lin-
earization technique in which we differentiate the output of the systems as many as times as
needed so that the input of the system appears in the last derivative[32, 1]. This technique is a
systematic way to linearize globally part of, or all, the dynamics of system [29]. The following
paragraphs explains how the new/synthetic input v is chosen in order to yield the following
Y(s) 1
= γ
V(s) s
where γ, the relative degree, is the last derivative of output so that the physical input appears in
the equation. If this order is less than the system order (n), then there will be internal dynamics
in the feedback linearized system. In cases when internal dynamics appears, the stability of
these dynamics should be also be considered. Here we consider a nonlinear system in the
42
following form:
ẋ = f (x) + g(x)u
y = h(x) (4.13)
where x(∈ Rn ) is the state vector, u(∈ Rm ) represents the control inputs, and y(∈ R p ) stands
for the outputs, f and g are smooth vector fields, and h is a smooth scalar function. A smooth
function is defined as an infinitely differentiable function. The control design process is to find
where α and β are smooth functions. This control law exactly linearizes the map between the
transformed input v and the output y and yields a linear system (linear from the synthetic input v
to the output y). The above idea can be implemented by successively differentiating the output
as:
where y(γ) represents the γth derivative of y, Lkf h(x) is called the Lie derivative of Lk−1
f h(x)
is given by
1
u= (−Lγf h + v) (4.16)
Lg Lγ−1
f h
−Lγf h(x)
α(x) =
Lg Lγ−1
f h(x)
1
β(x) = (4.17)
Lg Lγ−1
f h(x)
43
In order to facilitate the design block, assuming
We see that the above inversion-based control law has the capability to shape the output
response by simply designing the new control v to get the closed-loop linear system which
y(γ) = v (4.20)
Once linearization has been achieved, any further control objective such as pole placement
For the nonlinear quadcopter system under consideration in this section, in order to make
the system feedback linearizable, one may consider choosing x, y and z as the output variables.
It can easily be seen that u2 and u3 in (4.11), which are the control inputs representing the pitch
and roll angular accelerations of the vehicle, do not appear in the equation of the outputs. By
successively differentiating of equations of motion till the input terms appear, we can generate
the new control input of the system. It can be seen the new input terms appear in the fourth
" #
x (4)
= f (x) x + g(x) x1 g(x) x2 g(x) x3 u
" #
y(4) = f (x)y + g(x)y1 g(x)y2 g(x)y3 u
" #
z(4) = f (x)z + g(x)z1 g(x)z2 g(x)z3 u (4.21)
44
where
g(x)y2 = 0
45
where u is the control inputs which control the x, y and z. We choose u as:
−1
g(x) x1 g(x) x2 g(x) x2 − f (x) x + v1
u = g(x)y1 g(x)y2 g(x)y3 . − f (x)y + v2 (4.22)
g(x)z1 g(x)z2 g(x)z3 − f (x)z + v3
(4)
x v1
y = v2 (4.23)
(4)
(4)
z v3
(4) (3) (2)
v1 xd − k x1 e x − k x2 e x − k x3 e˙x − k x4 e x
v = y(4) − k e(3) − k e(2) − k e˙ − k e (4.24)
2 d y1 y y2 y y3 y y4 y
v z(4) − k e(3) − k e(2) − k e˙ − k e
3 d z1 z z2 z z3 z z4 x
[ky1 , ..., ky4 ] and [kz1 , ..., kz4 ] are gains, xd , yd and zd are desired outputs. From (4.24), the error
x − k x1 e x − k x2 e x − k x3 e˙x − k x4 e x = 0
e(4) (3) (2)
A PD controller is also design to control the yaw motion, and is given by:
where kψ1 and kψ2 are derivative and proportional gains respectively.
46
4.2.2 PD Based Tilting Angle Controller
Tilting rotor quadcopter is designed by using additional four servo motors attached to the
end of each arm that allow the rotors to tilt. This capability turns the vehicle into an over-
actuated system that potentially can track an arbitrary trajectory over time. During the flight,
as the non-linear control based on the proposed feedback linearization method provides the
amount of pitch and roll required to track an arbitrary trajectory, a PD controller is also designed
to allow the vehicle to either fly or hover with desired orientation or tilting. The relationship
between the tilt angles of the individual rotors, given by θides , i = 1, 2..4, and the reference pitch
θ1 1
h
0 1 0 2φ
h
θ2 0 1 0 1 2θ
= (4.27)
θ3 1 0 ∆φh
0 1
4
θ 0 1 0 1 ∆θ
h
where φh and θh are reference roll and pitch angles and ∆φh and ∆θh are orientation devia-
tions. A proportional-derivative controller is used to control the orientation deviation using the
where p, q (and t in the Equation below) are the components of angular velocities of the vehicle
in the body frame. The relationship between these components and derivatives of the roll, pitch
It may be noted that this PD controller is designed to desirably control both pitch and roll
angles. However, in this section, for design of feedback linearization method, we assumed the
quadcopter to be tilted just in roll direction. Hence, for the simulation studies carried out at
chapter 7, we set the reference pitch angle to be zero for this controller.
47
4.3 Stability Analysis
The inherently unstable tilting quadcopter dynamics described in (3.8) and (3.9) can be
written in state-space form: Ẋ(t) = f (X(t), U(t)) where U(t) and X(t) are input and state vectors.
x1 = x
x2 = ẋ1 = ẋ
x3 = y
x4 = ẋ3 = ẏ
x5 = z
x6 = ẋ5 = ż
x7 = φ
x8 = ẋ7 = φ̇
x9 = θ
x10 = ẋ9 = θ̇
x11 = ψ
x12 = ẋ11 = ψ̇
" #T
X = x ẋ y ẏ z ż φ φ̇ θ θ̇ ψ ψ̇
" #T
U = F1 F2 F3 F4 θ1 θ2
The state space model Ẋ(t) = f (X(t), U(t)) is not only non-linear but also highly compli-
cated due to tight coupling between different terms. In order to reduce the number of com-
plicated derivative terms involved in further dynamics, the small angle assumption has been
applied to differentiation described in (3.8) and (3.9). We have linearzed the system about an
operating hovering point while tilting along pitch or roll direction. The operating hovering
point Xe is achieved with the input (Ue ) such that f(Xe (t),Ue (t))=0. The linearized system is
given by:
48
Ẋ(t) = AX(t) + BU(t)
∂f ∂f
The linearization is carried out via calculating the Jacobian matrices as: A = ∂X
and B = ∂U
0 1 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 K1 0 K2 0 0 0
0 0 0 1 0 0 0 0 0 0 0 0
0 0 0 0 0 0 L1 0 0 0 0 0
0 0 0 0 0 1 0 0 0 0 0 0
0 0 0 0 0 0 M1 0 M2 0 0 0
A = ,
0 0 0 0 0 0 0 1 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 1 0 0
0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 1
0 0 0 0 0 0 0 0 0 0 0 0
49
0 0 0 0 0 0
A1 A2 A3 A4 A5 A6
0 0 0 0 0 0
B1 B2 B3 B4 B5 B6
0 0 0 0 0 0
C
1 C2 C3 C4 C5 C6
B =
(4.29)
0 0 0 0 0 0
D 0 D3 0 D5 0
1
0 0 0 0 0 0
0 E2 0 E4 0 E6
0 0 0 0 0 0
F1 F2 F3 F4 F5 F6
where:
1
K1 = (−F4 sinθ4 sinθcosφ + F2 sinθ2 sinθcosφ − F1 cosθ1 sinθsinφ
m
− F2 cosθ2 sinθsinφ − F3 cosθ2 sinθsinφ − F4 cosθ2 sinθsinφ)
1
K2 = (−F1 sinθ1 sinθ − F3 sinθ1 sinθ + F4 sinθ2cosθsinφ
m
+ F2 sinθ2 cosθsinφ + F1 cosθ1 cosθcosφ + F2 cosθ2 cosθcosφ
50
A1 = sinθ1 + cosθ1 sinθ, A2 = cosθ2 sinθ, A3 = sinθ1 + cosθ1 sinθ
F4 = lsinθ2 − kcosθ2 , F5 = lF1 cosθ1 + lF3 cosθ1 − kF1 sinθ1 − kF3 sinθ1
Once the system is linearzied, the controllability of the system in the vicinity of equilibrium
points can be analyzed using tools of linear system theory. We analyzed the system’s contolla-
bility for ranges of values on tiltilng angles along pitch or roll directions. As an example, we
θhdes = 200 .
51
0 1 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0.011 0 21.342 0 0 0
0 0 0 1 0 0 0 0 0 0 0 0
0 0 0 0 0 0 −10.340 0 0 0 0 0
0 0 0 0 0 1 0 0 0 0 0 0
0 0 0 0 0 0 3.762 0 −0.078 0 0 0
A = ,
0 0 0 0 0 0 0 1 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 1 0 0
0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 1
0 0 0 0 0 0 0 0 0 0 0 0
1 0 0 0 0 0 0 0 0 0 0 0
0 0 1 0 0 0 0 0 0 0 0 0
0 0 0 0 1 0 0 0 0 0 0 0
C =
0 0 0 0 0 0 1 0 0 0 0 0
0 0 0 0 0 0 0 0 1 0 0 0
0 0 0 0 0 0 0 0 0 0 1 0
52
0 0 0 0 0 0
0.0314 0.0028 0.0314 0.0028 11.8610 −0.0114
0 0 0 0 0 0
0 0.3421 0 0.3421 0 10.2456
0 0 0 0 0 0
1.0000 0.9397 1.0000 0.9397 −0.0335 0
B =
0
0 0 0 0 0
−0.2500 0 0.2500 0 −0.0400 0
0 0 0 0 0 0
0 −0.2281 0 0.2281 0 −0.9028
0 0 0 0 0 0
0.0250 0.0667 0.0250 0.0667 0.5000 0.4562
The matrices A and B were used to determine the controllability of the system and was
found to be controllable. This means that a feedback control law U(t) = −K f d X(t) can be
designed to stabilize and control the system via pole placement method where K f d is a 6 × 12
53
Chapter 5
5.1 Introduction
In this chapter, stability and control of tilting-rotor quadcopter is presented upon failure
of one propeller during flight. The tilting rotor quadcopter provides advantage in terms of
additional stable configurations. On failure of one propeller, the quadcopter has a tendency of
spinning about the primary axis fixed to the vehicle as an outcome of the asymmetry about the
and it is capable of handling a propeller failure, thus making it a fault tolerant system. In this
chapter, a dynamic model of tilting-rotor quadcopter with one propeller failure is derived and a
controller is designed to achieve hovering and navigation capability. The simulation results of
Multicopters with six or more propellers are also popular as the vehicle is able to maintain
normal flight if one of the propellers fails [48]. But multicopters are costly as compared to
the quadcopters while applications are the same. VTOL UAVs are finding more and more
applications in civilian domain and this changing scenario demands new rules and regulations in
future[63, 89]. System failures are inevitable during flight of UAVs. Propeller or motor failure
is one of the most common failure in case of quadcopters[6]. Currently, the commercial solution
available to deal with propeller failure is emergency parachute which assists in emergency
The operational scenario of quadcopters requires the design of controllers capable of fault
detection, isolation, and diagnosis [3]. Once the failure occurs, the system must be capable of
54
maintaining the stability of the system and complete the mission without much compromise in
system performance. In passive fault tolerant control system (PFTCS) the control algorithm is
designed to achieve a given objective in healthy or faulty situation without changing its control
law [93], whereas In active fault tolerant control system (AFTCS), to preserve the ability of sys-
tem to achieve the objective, the control law is changed according to fault situation [8, 42]. The
fault diagnosis and identification (FDI) block, also termed as diagnosis unit, consists residual
generator and residual evaluation sub-units. A residual is generated by comparing the process
output and the model output, if the residual differs from zero. The residual evaluation compares
it to a threshold to decide and indicate fault. Based on the diagnosis result the reconfiguration
block has to adapt the controller in such a way that the new controller is able to cope with the
faulty process.
Fault Detection and Isolation (FDI) system for actuator faults for an hexacopter vehicle has
been presented in [20] . A diagnostic Thau observer is applied to the hexacopter nonlinear
model to generate residual signals. In the fault-free case, residuals are close to zero, while in
case of a faulty actuator, the value of residuals and fault is detected. Further, Fault isolation
is realized by exploiting the mathematical model of the hexacopter. By quickly detecting the
fault, the control law can be modified to satisfy the closed-loop requirements of the system and
In [48] the control strategy is presented using periodic solutions for a quadcopter experi-
encing one, two opposite, or three complete rotor failures. The strategy employed is to define
an axis, fixed with respect to the vehicle body, and have the vehicle rotate freely about this axis.
By tilting this axis, and varying the total amount of thrust produced, the vehicles position can
be controlled.
Emergency landing procedure of quadcopter has been presented in [38, 41, 24] by using
PID and Backstepping control approach respectively. The strategy is to switch off the propeller
aligned on the same quadcopter axis of the failed propeller. This action converts the quad-
copter configuration into a birotor aerial vehicle. The UAV becomes free to spin in yaw axis
while controlling the remaining attitudes of the UAV and then emergency landing procedure is
exercised.
55
The tilting-rotor quadcopter is an over-actuated [51, 52, 82] form of a traditional quadcopter
and it is capable of handling a propeller failure, thus making it a fault tolerant system. A
robust, fault tolerant control law and redundant mechanical design of the quadcopter can ensure
safe handling of the quadcopter even after the propeller failure. In this chapter, the tilt rotor
mechanism and PD control of the quadcopter have been used to stabilize the quadcopter after
the propeller failure and thus control all states of the UAV.
Fault Detection and Isolation (FDI) system for motor failure is a very important aspect of
fault tolerant control for quadcopters. Quadcopters belong to the class of very fragile aircraft
and if a motor failure occurs, it leads to highly unstable system dynamics. On failure of one
propeller, the quadcopter has a tendency of spinning about the primary axis fixed to the vehicle
as an outcome of the asymmetry about the yaw axis. The second major asymmetry is created
in the roll or pitch plane depending on the corresponding motor failure. If any one of motor
1 or 3 fails, the asymmetry will be in pitch and yaw plane whereas if motor 2 or 4 fails there
would be a roll and yaw asymmetry. A robust Fault Detection and Isolation (FDI) system
can minimize the reaction time for control system reconfiguration and improve the efficiency
of fault tolerant control significantly. As such, this mechanism plays a key role in FTC. FDI
can be implemented with a current sensor that can be used to monitor the amount of current
supplied to each quadcopter motor. The signal from this sensor can be used to take identify the
fault and take further decision for control system reconfiguration. These current sensors fall in
Unlike traditional quadcopter models, which have only four rotary propellers as the vehi-
cle’s inputs, in tilting rotor quadcopters, there are four more servo motors attached to the each
arm that adds one degree of freedom to each of the propellers, resulting in the tilting motions
along their axes. The equation of motion of a tilting rotor quadcopter has been discussed in
56
previous sections. In this section, the equations of motion of a tilting rotor quadcopter with one
When all the propellers of the tilt-rotor quadcopter are working then it yields a stable config-
uration as a result of symmetry of forces and moments. Assuming that one propeller/motor fails
during hovering flight of quadcopter which is located in the pitch plane. Then, the quadcopter
would possess three working propellers and one failed propeller. Once the failure occurs, the
UAV will experience asymmetry about the yaw axis because of M1 , M3 , M4 moments of work-
ing propellers while M2 = 0. Another asymmetry would occur in pitch plane as F2 = 0 and F4
would still have some magnitude. The equations of motion can be modified by putting F2 and
M2 equal to zero.
Once again by using rotational matrix in (3.1), equations of motion in world-frame can be
written as:
+ F4 cθ4 sψsφ − C1 ẋ
− F4 cθ4 cψsφ − C2 ẏ
− mg − C3 ż (5.1)
It should be noted that F2 terms have vanished from the equations which will result in asym-
57
metry because of one propeller failure.
+ M4 0
+ (M1 0 + M3 0 )
where M 0i , (i = 1, 2, 3, 4) are the same tilting moments which are created by the four servo
motors attached to the end of each arm to cause a tilt angle. The absence of F2 , M2 should be
noted in pitch and yaw acceleration equations. The components of rotor moment M1 , M3 , M4
would not produce a symmetrical outcome which represent unstable dynamics of quadcopter
upon propeller failure. The available inputs to stabilize and control this system are angular
given by Equations (5.1) and (5.2), the quadcopter can be stabilized in yaw and pitch plane if
fourth rotor is tilted by an angle θ4 such that θ4 = c−1 [ω24 /(ω21 cθ1 + ω23 cθ3 )].
Figure 5-1: Free body diagram of tilt-rotor quadcopter upon propeller failure
Proof: When propeller failure occurs the dynamics of the quadcopter are highly non-linear.
58
Thus, we ignore the drag forces and moments generated because of rotor tilt for simplification.
This assumption simplifies the angular acceleration equations and the equations for pitch and
If the quadcopter has to be stabilized in pitch and yaw plane, θ̈, ψ̈ should be zero. Thus,
Substituting the value lF4 in equation (5.4) and rearranging the equation:
The above expression can be re-written in terms of angular velocity by using equation (3.3) :
This condition should hold for attaining a stable configuration after one propeller failure in the
tilt-rotor quadcopter. Otherwise, the system can not be stabilized or controlled. In fact, once
59
the system is stabilized minor deviation in angular speeds of propellers and rotor tilt angle can
Theorem-IV: Once propeller failure occurs, the quadcopter can hold a certain altitude if the
0
ω1 = ω1 + ω2 /3
0
ω3 = ω3 + ω2 /3
0
ω4 = ω4 + ω2 /3 (5.8)
which means:
ωnew
h = ωh + ω2 /3 (5.9)
Proof: ω2 represents the angular speed of the second propeller at the instant of failure, this
will result in the loss of altitude but the angular speed of three remaining propellers can be
increased by a factor of ω2 /3 in order to compensate for the loss. On the other hand, an extra
compensation component ω4 /cθ4 should come in the equation of fourth rotor to overcome the
tilt effect of the rotor. Thus, angular speed of fourth rotor will be higher as compared to angular
speed of first and third rotor. ω1 , ω3 , ω4 are the increased angular speeds of the propellers in
0 0 0
Future, the new angular speeds must satisfy equation (5.7) in order to yield a stable config-
0 0 0
θ4 = c−1 [ω42 /(ω12 cθ1 + ω32 cθ3 )] (5.10)
60
5.4 Controller Design
In this section, the control strategy of the tilting rotor quadcopter in a case of motor failure
during the flight is presented. Two PD controllers are used due to compensate the unbalance
moments created by an odd number of propellers, and also to stabilize vehicle’s orientation
and make it functional to continue its mission without crash. The vehicle originally has eight
independent inputs which includes four speed of propellers and four tilted angle of each motor
about its axis. In the case of motor failure, two inputs are automatically out of equations. To
make the vehicle compensate the moments of the vehicle, not only the speed of the remaining
propellers needs to be controlled individually, but also the tilted mechanism needs to be set
in a way that compensates the moment from the breakdown motor. In this work, its assumed
that motor two is the one that stopped working during the flight. It should be noted that the
measurement sensor needs to report the failure immediately. Referring to the theorems, the
tilting angle of motor 1 and motor 3 needs to be set at the same orientation and the tilted angle
To start compensating the unbalanced moment situation after the failure, first, getting back
to hovering is necessary. Then, the orientation of the vehicle to a specific pitch or roll angle is
obtained. In [46], the relationship between the rotational speeds of the motors and the deviation
of the orientations from nominal vectors for hovering and navigation is described in detail
for conventional quadcopter. Based on the new dynamics equation of the vehicle with three
propellers and one tilted servo motor (motors 1 and 3 are assumed to be level), the following
ω0
1
M x Lk f ω1 −Lk f ω3
B 0 0
0 0 0
ω
3
My = 0 Lk f ω4 0 0
0
(5.11)
B
0
ω
M B 0 4
z 0 0 1
θ4
where M xB , MyB and MzB are torque components separated in body frame. It needs to be men-
tioned that these equations are obtained from linearization of equation (5.2) around its nominal
61
hover states while first and third servo motors are assumed not to be tilted. The rotational speed
on each individual motor and the tilted angle of the rear motor are calculated as:
ω1 1 −1 0 0 ωh + ∆ω f
des new
ωdes 1 1 0 0 ∆ω
3 φ
=
(5.12)
ωdes 1 0 1 0 ∆ωθ
4
θ4 θ4 + ∆θ4
des
0 0 0 1
where ωdes
i , (i = 1, 3, 4) are the desired angular velocities of the respective rotors. θ4 is the tilted
angle that needs to be hold for motor 4 to attaining a stable configuration and is calculated in
derivative laws are used to control ∆ωφ , ∆ωθ , ∆θ4 and ∆ω f which are deviations that result into
forces/moments causing roll, pitch, yaw, and a net force along the zB axis, respectively, which
where p, q and t are the component of angular velocities of the vehicle in the body frame.
During the flight of a tilting quadcopter, the orientation of the vehicle needs to be set level.
This can be obtained by linearizing the equation of motion that correspond to the nominal
equilibrium hovering configuration with the reference pitch or roll angles. The change of the
pitch or roll angles are supposed to be small during flight. By linearizing Equation (3.8) about
these nominal hovering states, desired pitch and roll angles to cause the motion can be derived
62
where θdes and φdes are the desired pitch and roll to be added to the nominal hover states to
move the vehicle to desired trajectory ri,T , the command acceleration, r̈ides is calculated from
where ri and ri,T (i = 1, 2, 3) are the 3-dimensional position of the quad-rotor and desired
The dynamic model of the vehicle with one motor failure can be described with the differ-
ential equations (5.1) and (5.2). This inherently unstable tilting quadcopter dynamics can be
written in state-space form: Ẋ(t) = f (X(t), U(t)) where U(t) and X(t) are input and state vectors
[10].
x1 = x
x2 = ẋ1 = ẋ
x3 = y
x4 = ẋ3 = ẏ
x5 = z
x6 = ẋ5 = ż
x7 = φ
x8 = ẋ7 = φ̇
x9 = θ
x10 = ẋ9 = θ̇
x11 = ψ
x12 = ẋ11 = ψ̇
63
" #T
X = x ẋ y ẏ z ż φ φ̇ θ θ̇ ψ ψ̇
" #T
U = F1 F3 F4 θ4
Since the quadcopter will not be used for acrobatic maneuvers after the motor failure, the
highly nonlinear and also complicated dynamics, can be simplified with the small angle as-
sumption [87] to cover hovering and moving around with small deviation in orientation. In
order to reduce the number of complicated derivative terms involved in further dynamics, the
small angle assumption has been applied to differentiation described in (5.1) and (5.2). We have
These operating hovering point Xe is achieved with the input (Ue ) such that
where:
∂ fi (X0 , U0 )
ai, j =
∂X j
∂ fi (X0 , U0 )
bi, j =
∂U j
The linearization is carried out via calculating the Jacobian matrices [40] yields:
64
0 1 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 A2,7 0 A2,9 0 0 0
0 0 0 1 0 0 0 0 0 0 0 0
0 0 0 0 0 0 A4,7 0 0 0 0 0
0 0 0 0 0 1 0 0 0 0 0 0
0 0 0 0 0 0 A6,7 0 A6,9 0 0 0
A = ,
0 0 0 0 0 0 0 1 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 1 0 0
0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 1
0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0
A2,1 A2,2 A2,3 A2,4
0 0 0 0
A4,1 A4,2 A4,3 A4,4
0 0 0 0
A
6,1 A6,2 A6,3 A6,4
B = (5.17)
0 0 0 0
A
8,1 A 8,2 0 0
0 0 0 0
0 0 A10,3 A10,4
0 0 0 0
A12,1 A12,2 A12,3 A12,4
65
where:
1
A2,7 = (−F4 sinθ4 sinθcosφ − F1 cosθ1 sinθsinφ − F3 cosθ3 sinθsinφ
m
− −F4 cosθ4 sinθsinφ)
1
A2,9 = (−F1 sinθ1 sinθ + F3 sinθ3 sinθ − F4 sinθ4cosθsinφ
m
+ F1 cosθ1 cosθcosφ + F3 cosθ1 cosθcosφ − F4 cosθ4 cosθcosφ)
1
A4,7 = (F4 sinθ4 sinφ − F1 cosθ1 cosφ + F3 cosθ3 cosφ − F4 cosθ4 cosφ)
m
A6,7 = F4 sinθ4 cosθcosφ − F1 cosθ1 cosθsinφ − F3 cosθ3 cosθsinφ − F4 cosθ4 cosθsinφ)
A4,1 = −cosθ1 sinφ, A4,2 = −cosθ3 cosφ, A4,3 = −sinθ4 cosφ − cosθ4 sinφ
A6,1 = −sinθ1 sinθ + cosθ1 cosθcosφ, A6,2 = sinθ3 sinθ + cosθ3 cosθcosφ
A6,3 = −sinθ4 cosθsinφ + cosθ4 cosθcosφ, A6,4 = −F4 cosθ4 cosθsinφ − F4 sinθ4 cosθcosφ
66
Once the system is linearized, the controllability of the system in the vicinity of equilibrium
0 1 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 −4.1241 0 8.249 0 0 0
0 0 0 1 0 0 0 0 0 0 0 0
0 0 0 0 0 0 −2.751 0 0 0 0 0
0 0 0 0 0 1 0 0 0 0 0 0
0 0 0 0 0 0 4.763 0 0 0 0 0
A = ,
0 0 0 0 0 0 0 1 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 1 0 0
0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 1
0 0 0 0 0 0 0 0 0 0 0 0
1 0 0 0 0 0 0 0 0 0 0 0
0 0 1 0 0 0 0 0 0 0 0 0
0 0 0 0 1 0 0 0 0 0 0 0
C =
0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0
67
0 0 0 0
0 0 0 0
0 0 0 0
0 −1.000 −0.865 −2.751
0 0 0 0
1.000 1.000 0.500 −4.763
B =
0
0 0 0
−0.250 0.250 0 0
0 0 0 0
0 0 0.142 −1.136
0 0 0 0
0.020 0.020 0.026 1.878
A feedback control law [58] U(t) = −K f d X(t) can be designed to stabilize and control the
system via pole placement method where K f d is a 4 × 12 feedback control gain matrix as below:
T
0.177 −0.265 0.326 0.007
1.517 −2.079 2.680 0.034
−0.345 0.320 −0.499 −0.058
−1.183 1.028 −2.059 −0.247
−0.675 0.244 −0.091 −0.082
−4.799 1.442 −0.324 −0.609
=
Kfd
−54.857 29.104 −23.450 −4.416
−48.255 17.135 −11.818 −3.983
22.162 −35.742 50.339 −0.500
34.199 −26.875 31.562 1.374
0.785 −0.790 0.870 0.099
3.298 −3.746 4.039 0.389
68
Figures (5-2) and (5-3) show how position and orientation of quadcopter with one failed motor
remaining stable.
2
X
Y
1.5 Z
1
meter
0.5
-0.5
0 10 20 30 40 50
time(sec)
0.35
φ
0.3
θ
ψ
0.25
0.2
rad
0.15
0.1
0.05
-0.05
0 10 20 30 40 50
time(sec)
69
Chapter 6
Hardware design
The development of a proposed control techniques for tilting quadcopter requires the devel-
opment of an adequate platform for the preliminary experiments. The primary consideration
for the prototype design was to make the quadcopter small and lightweight so that it was able to
carry an extra component required for the tilting mechanism during the flights. The initial con-
figuration and concept of the tilting quadcoter are presented in Figure (6-1). As the fabricating
process was a senior design project, the cost of the vehicle was kept as low as possible.
The structural design of the tilting quadcopter can be divided into two parts: the central
body where all avionics and components are placed, and the tilting mechanism.
70
6.1.1 Central body
As the most important mechanical components of the tilting mechanism are located at the
center area, this part not only needs to have enough space for all components, but also needs
to maintain its symmetry requirement. The central body or core area needs to have space for
• Communication Hardware
• Autopilot
• Bearings, and
• Servo motors.
There are two prototypes which have been designed and fabricated. The core part in the
first prototype is made using two sheets of flame-retardant Garolite which is connected by
plastic bolts and spacer. The flame-retardant Garolite offers excellent strength, low water ab-
sorption and good electrical insulating quality in both humid and dry conditions with maximum
retardant resin. The hardness also meets Rockwell M110-M115 which is standard hardness
test. The reason to use two sheets is to have all bearings, servo motors and electronic parts in
between. Four additional polycarbonate round tubes are also attached to the frame in order to
make four arms for the quadcopter. Polycarbonate round tube is a cost effective material with
weight, has excellent impact strength, heat resistance to 2500 F and most importantly it comes
in a round shape which provides the tilting mechanism with more degree of freedom to rotate.
The most important features of the tilting quadcopter is the fact that each arm can ro-
tate about its axis independently. In order to have this independency, four additional inputs
71
are needed to rotate each arm to ensure the tilting mechanism. The servos (Figure (6-2)) are
mounted on top of the lower sheet of the central core. Each arm is directly linked to the servo
motor using gears. The gear is hooked around the tube in the middle by two bearings. All eight
bearings (two for each arm) mount between upper and lower sheets. This unique design allows
the arms to rotate separately with any desired angle without mechanical constraint. Figure (6-3)
Table (6.1) and (6.2) summarize the specifications and the masses of the vehicle’s compo-
without propellers
Autopilot Pixhawk
Figure 6-4: The real model of the first model of tilting Quadcopter.
After ten test flights, due to weight of the vehicle, the motors were not strong enough to
handle the tests and the experimental data was not acceptable. To reduce the weight and also
make the tilt mechanism more agile, second prototype was designed in Autodesk Inventor and
SolidWorks and fabricated by our group. For the second prototype, the 3D robotics framework
was used as the body and four separate tilt mechanisms were designed and built by using a
73
3D printer. The printed 3D mechanism was mounted at the very end of each arm. The parts
were made in a way that perfectly fit the arms to avoid any slinging. The electronics, power,
propellers, motors and the auto pilot are same as the first prototype.
Motors 80 4 320.0
Servos 45 4 180.0
Prop adapters 8 4 32
Plastic Bolts 6 16 96
ESC 15 4 60
bearings 35 8 280
Landing gear 50 1 50
Controller 38 1 38
Battery Holder 45 1 45
Arms 35 4 140
Total 2291
This vehicle also has eight control inputs which are used for rotating the four propellers
and tilting mechanism for the arms. It weights 1kg less than the first prototype. The diameter
also does not exceed 55cm. To tilt the motors around the arms, a tilting mechanism has been
located at the end of each arms. Each tilting mechanism consist of three separate parts: i) the
servo motor holder ii) the plate holder iii) and the motor plate which is mounted on the base
There are some similar mechanisms available in the market, but the main reason that makes
74
our design more reliable, is the way the servo is connected to the tilting part. In available
versions of the tilting mechanism, the servo is screwed to the mechanism with the same screw
which holds the tilting parts together. The problem comes when the screw is tightened enough
to attach the servo to the tilting part to avoid any sliding. This also pushes the two separate
mechanisms towards each other and makes it harder to tilt. While more force is needed to be
able to make the pressed part to rotate, an extra force which comes from the servo motor, makes
the cog loose. The looseness between the servo and the tilting part contributes to a delay when
the command is received the motor starts tilting. This delay adds the nonlinear parts to the
The currently designed tilting mechanism has an extra screw which is placed between the
tilting mechanism and the servo cog. By having this additional screw, the servo cog can be
fastened to the tilting mechanism as much as needed, while the two tilting mechanisms are
attached to each other with separate screw. With this design, not only there is no friction in
tilting mechanism, the servo would never get loose. Figure (6-5) shows the transparent CAD
model of the tilting mechanism. Figure (6-6) shows how the tilting mechanism is mounted at
the end of quadcopter’s arm and how the servo, tilting mechanism and the motor are connected.
75
Figure 6-6: The CAD model of tilting mechanism mounted on the arm with the servo and the
motor
Figure 6-7: The CAD and the actual model of the prototype.
As the servo is directly connected to the tilting mechanism with no gears in between. The
smaller and lighter servo which produces the lower torque compared to the previous design can
also be used. This configuration not only eliminates external gears which were used before,
but also reduces the weight of the servo by half. One of the drawback of this mechanism as
compared to the previous one, is the freedom of tilting around the axis. In previous design
there was not any mechanical limitation for the tilting mechanism. Although the previous
design could rotate around the arms up to 360 degrees, it never was our concern. However, this
76
tilting mechanisms is limited to 60 degrees which meets the criteria of the experiments.
6.2.1 Motors
A multi-rotor flying vehicle is more efficient when it is lighter. One of the important criteria
to choose a suitable motor is to use the lightest possible motor which can provide at least
twice as much thrust to lift the vehicle and has the best response for control system in difficult
flight conditions. Brush-less DC motors afford better efficiency and power density compared to
brushed DC motor [31]. Due to higher power density, controllability, minimum requirements
for maintenance, compact size and light weight [60] they have become increasingly popular
As a result of these advantages, the 850 KV Brush-less DC (BLDC) motor was selected
for the tilting quadcopter. It is also used by most of the 3D Robotics multi-rotors vehicles. It
delivers a very good level of the thrust compared to even bigger size of counterpart 850 KV
motors.
Most brush-less DC motors have three terminals as shown in Figure (6-8). While these
three terminals are connected to stator, the permanent magnets are placed on the rotor such that
the poles are facing the stator. An Electronic Speed Control (ESC) is used to send the command
to control the rotation of the motor. The power output from ESC by using four cells LiPo and
6.2.2 Batteries
Lithium Polymer (LiPo) batteries are currently the preferred power sources for most light-
weight, high-current, high-capacity power storage [7]. They offer high energy-storage/weight
ratio and high discharge rate [88]. These batteries use normal lithium ion chemistry with the
polymer separators to provide high discharge rates. It comes with different cell numbers. Each
cell provide 3.7 Volts with internal resistance of approximately 0.03 Ohm [4].
6.3 Avionics
A Control Board needs to be selected that would maintain list of requirements to be able
to have control over position and orientation. As there is no control board for tilt mechanism
in the market, in order to apply the proposed control strategy, it needs not only to be an open
source, but also is required to be able to communicate with an out source computing devices.
Furthermore three axis gyroscope and three axis accelerometers are also need to be included.
Another criteria which was taken into the account to choose the control board was the reason-
able price, the size and the weight. All these reasons led us to use PixHawk autopilot [44].
PixHawk is armed with advanced 32 bit ARM processor, micro SD card for logging, Integrated
backup systems and 14 PWM servo outputs which are ideal for our vehicle to be connected to
additional tilting servo motors [45]. Another specification that made us to use PixHawk was
78
the way it could communicate with Robot Operating System (ROS) which is the main coding
environment to apply the control techniques [66]. ROS is a set of software frameworks for
robot software development. ROS also provides low-level device control and message pass-
ing between processes which allows to share data across multiple and commonly specialized
processes [16].
6.3.2 Communications
MAVLink (Micro Air Vehicle Link) is a very lightweight protocol for communicating with
It is mostly used for communication between ground station and small unmanned vehicles.
It can be used to transmit all flight data including altitude, attitude, GPS position, air speed,
battery status, way points etc[84, 85, 83]. The transmitted data has the packet structure. The
payload from the packets are called as a MAVLink message which is identifiable by the ID for
each message. The stream of bytes that can be encoded by the ground station can be sent via
MAVROS is also the package in ROS environment that provides communication driver
with MAVLink communication protocol for various autopilots including PixHawk. Not only
all flight data is available in ROS, but also the new commands which are calculated through
79
Chapter 7
Results
To validate the presented dynamic model and the control method, numerical simulations of
the tilting rotor quadcopter were carried out using the MATLAB. The discretized versions of
the dynamic and the controller equations are solved by the Euler method. Here, we provide
the results from one of the simulation scenarios studied. In this scenario, the vehicle’s initial
position was (0.1, 0.8, 0). The final position was set to (0.8, 0.3, 1.5). In this study the SI unit
system is used. In this scenario, the desired pitch or roll angles were modified during the flight
so that both of these angles were simultaneously controlled. Figure (7-1) shows the reference
pitch and roll angle. It can be seen that for time t = 0 sec to t = 1 sec, the reference pitch and
roll angles are zero. At t = 1 sec, the reference pitch angle increases from 00 and reaches the
value of 180 at t = 4 sec and stays with the same until the end of trajectory. At t = 4 sec, the
reference roll angle is also increased from 00 and reaches the value of 120 at t = 7 sec.
80
20
Pitch (degree)
15
10
0
0 1 2 3 4 5 6 7 8 9 10
time(sec)
15
Roll (degree)
10
0
0 1 2 3 4 5 6 7 8 9 10
time(sec)
The procedure to accomplish the flight simulation is to have the vehicle take-off from an
initial point vertically till the desired height, and then steer to the destination point with the
horizontal flight. During flight, the orientation of the vehicle is supposed to change according
to reference inputs without losing the height. The quadcopter trajectory in the three dimensional
space from the initial point to the desired destination is shown in Figure (7-2).
1.5
Z(m)
0.5
0
1
0.5
Y(m) 0.6 0.8
0.2 0.4
0 0
X(m)
81
Figure (7-3) shows the actual change in the pitch, roll, and yaw angles of the tilting rotor
quadcopter during the flight. It can be seen that the change in the yaw angle is close to zero
while the actual pitch and roll angles closely follow the reference values.
20
θ(degree)
−20
0 1 2 3 4 5 6 7 8 9 10
time(sec)
20
φ(degree)
−20
0 1 2 3 4 5 6 7 8 9 10
time)sec)
−4
x 10
4
ψ(degree)
0
0 1 2 3 4 5 6 7 8 9 10
time(sec)
40 30
20
θ1 (degree)
θ2 (degree)
20
10
0
0
−10
−20 −20
0 2 4 6 8 10 0 2 4 6 8 10
time(sec) time(sec)
20 20
10
θ3 (degree)
θ4 (degree)
0
0
−10
−20
−20
−40 −30
0 2 4 6 8 10 0 2 4 6 8 10
time(sec) time(sec)
Figure (7-4) shows how four servomotors modulate the angle of each arm to follow the
referenced orientation commands during the flight. Figure (7-5) shows how the speed of four
motors changes during the flight to track the trajectory and maintain the height of the vehicle.
Figure (7-6) shows the enlarged view of a portion of the Figure (7-6). It can be seen from Figure
82
(7-6) that the angular velocities of rotors stabilize close to t = 1sec (till when the reference tilts
in pitch and roll are zero). At t = 1sec, the reference pitch angle is commanded to gradually
increase to reach a value of 180 at t = 4sec. Corresponding to this, the individual rotor speeds
can be seen to increase from t = 1sec to t = 4sec. At t = 4sec, the reference roll angle is
is further increase in the rotational speeds of the rotors. This increase in motor speed can be
(3.7) (for non-tilted configuration), the theory predicts the need of more rotor speed in tilted
configuration so that the vertical component of force still balances the weight in the tilted
configuration. The simulation results shown in Figure (7-6) just confirms the theory.
150 150
100 100
ω1(rad/s)
ω2(rad/s)
50 50
0 0
0 2 4 6 8 10 0 2 4 6 8 10
time(sec) time(sec)
150 150
100 100
ω3(rad/s)
ω4(rad/s)
50 50
0 0
0 2 4 6 8 10 0 2 4 6 8 10
time(sec) time(sec)
83
115 115
ω1(rad/s)
ω2(rad/s)
110 110
105 105
100 100
2 4 6 8 10 2 4 6 8 10
time(sec) time(sec)
115 115
ω3(rad/s)
ω4(rad/s)
110 110
105 105
100 100
2 4 6 8 10 2 4 6 8 10
time(sec) time(sec)
In order to verify the performance of the proposed feedback linearization based control
method of the tilting-rotor quadcopter, numerous numerical simulations were carried out in
MATLAB/Simulink environment. Here, we present the results from one of the simulations.
The simulation comprises of the UAV taking off from the initial position located at (5.0 , 0.0
,0.0 ) and reaching the desired destination located (4.5, 10.0, 10.0), via passing the waypoint
located at (3.0, 4.0, 5.0). Also, during the flight, the orientation of the vehicle is supposed to
change according to the reference input (in this case, the desired roll angle) without deviating
from the desired trajectory. The reference roll angle is set as follows. During the flight, at
t = 5 sec, the reference roll angle is commanded to gradually increase from 0o to reach a value
of 12o at t = 10 sec. The quadcopter is then supposed to move towards the destination with
the commanded roll angle (refer to the bottom plot of Figure (7-9)). The quadcopter trajectory
obtained from the numerical simulation in the three-dimensional space from the initial point to
the waypoint and then to the final position is shown in Figure (7-7).
84
10
( 4.5, 10.0, 10.0 )
8
z ( meter)
6
0
10
( 3.0, 4.0, 5.0 )
8
2 6
5.5
5
4.5
y ( meter) 0 3.5
4
3
x ( meter)
Figure (7-8) shows the position and orientation of the vehicle during the flight. Figure (7-9)
shows the comparison between the reference roll and the actual roll angle the the vehicle.
10
10
8
8
z (meter)
y (meter)
6
6
4 4
2 2
0 0
0 5 10 15 20 2 3 4 5 6
time(sec) x (meter)
10 0.1
5 0.05
ψ (degree)
θ (degree)
0 0
−5 −0.05
−10 −0.1
0 5 10 15 20 0 5 10 15 20
time(sec) time(sec)
Figure 7-8: Position and orientation of the quadcopter: altitude vs. time (top left), x-position
vs. y-position (top right), pitch vs. time (bottom left), and yaw vs. time (bottom right).
.
85
15
φ (degree)
10
0 2 4 6 8 10 12 14 16 18 20
time(sec)
15
φh (degree)
10
0 2 4 6 8 10 12 14 16 18 20
time(sec)
Figure 7-9: The reference roll (bottom) and actual roll angle (top) during the flight.
15 15
10 10
u2 (rad/sec2)
u1 (m/sec2)
5 5
0 0
−5 −5
0 5 10 15 20 0 5 10 15 20
time(sec) time(sec)
−3
x 10
15 15
10
u3 (rad/sec2)
u4 (rad/sec )
10
2
5
5
0
0
−5
0 5 10 15 20 0 5 10 15 20
time(sec) time(sec)
and u4 as the normalized total lift forces and control inputs for roll, pitch, and yaw respectively,
Figure (7-10) shows that the zero dynamics of the controller are stable ( as the relative degree
86
is smaller than the order of the system).
To validate the presented dynamic model in the case of motor failure and the proposed con-
trol technique, numerical simulations of the tilting rotor quadcopter were carried out using the
MATLAB. The discretized versions of the dynamic and the controller equations are solved by
the Euler method. We have completed two different scenarios in order to evaluate the vehicle’s
response in the case of motor failure in following a trajectory as well as hover flight. In both
The first simulation is the hovering task in one spot with motor failure after stable hover
flight. This scenario shows the performance of the controller and highlights the position control
with motor failure. In the first scenario, mission was started by taking off and hovering in the
fixed altitude and in one spot. The initial take of point is located at (0.2, 0.0, 0.0). Figure
(7-11) shows the 3 dimensional path of the flight. As it can be seen, the vehicle is flying in the
neighborhood of the desired spot. The flight has the error in the range of 0.4 meters to hover
87
15
10
position Z
0
0.6
0.4 0.6
0.2 0.4
0.2
0 0
position Y -0.2 -0.2 position X
In this scenario the vehicle flew for 200 seconds. Figure (7-12) shows the position of vehicle
in each individual axis. At time t = 20 sec, when motor number 2 stops working, the altitude
of the quadcopter drops and after very short time of adjusting, it maintained its altitude to the
end of the flight. The vehicle also has small amount of movement along X and Y axis.
Figure (7-13) shows how the other motors increase the rotational speed in order to com-
pensate the failed motor force to maintain the altitude. Although the rotational speed of the
motors is increased, but the speed limitation of each motor, will not let the vehicle to maintain
88
1
X(m)
0
-1
0 50 100 150 200
time(sec)
0.5
Y(m)
-0.5
0 50 100 150 200
time(sec)
20
Z(m)
10
0
0 50 100 150 200
time(sec)
ω 2 (rad/s)
100 100
50 50
0 0
0 100 200 0 100 200
time(sec) time(sec)
ω 3 (rad/s)
ω 4 (rad/s)
100 100
50 50
0 0
0 100 200 0 100 200
time(sec) time(sec)
89
×10 -3
1
φ(rad)
0
-1
0 50 100 150 200
time(sec)
0.01
θ(rad)
-0.01
0 50 100 150 200
time(sec)
0.4
ψ(rad)
0.2
0
0 50 100 150 200
time(sec)
Figure (7-14) shows how orientation of the vehicle change after the motor failure. As it can
be seen from the Figure (7-14), the yaw angle changes right after the motor failure and remains
in a constant angle. It can easily be notified that the vehicle is not spinning around Z axis after
Here, we provide the results from the second simulation we studied. In this scenario, the
vehicle’s initial position was (0.0, 0.0, 0.0). The final position was set to (30.0, 10.0, 10.0). In
this scenario, It can be seen that for time t = 0 sec to t = 10 sec, the quadcopter flew with all
propellers working and it reached its designated altitude at 10 meters. at t = 10 sec, one the
motors stopped working and the proposed control technique were applied immediately. During
the simulation, it is assumed that right after the motor failure, the control system can take over
The quadcopter trajectory in the three dimensional space from the initial point to the desired
90
15
10
position Z
0
15
10
40
5 30
20
0 10
position Y 0
position X
Figure (7-16) shows the position of vehicle in each individual axis. At time t − 10 sec, when
motor number 2 stops working, the altitude of the quadcopter drops by 3 meters and after very
short time of adjusting, it maintained its altitude to the end of the flight.
40
X(m)
20
0
0 10 20 30 40 50 60
time(sec)
20
Y(m)
10
0
0 10 20 30 40 50 60
time(sec)
20
Z(m)
10
0
0 10 20 30 40 50 60
time(sec)
91
Figure (7-17) shows how the other motors behaved after motor 2 failure. As it can be seen
from the Figure (7-17), to compensate the lost force from motor 2, all working motors speed
up. Although the rotational speed of the motors is increased, but still the altitude is decreased
to 7 meters due to limitation of the amount of rotational speed each motor can provide.
ω 1 (rad/s)
ω 2 (rad/s)
100 100
50 50
0 0
0 20 40 60 0 20 40 60
time(sec) time(sec)
ω 3 (rad/s)
ω 4 (rad/s)
100 100
50 50
0 0
0 20 40 60 0 20 40 60
time(sec) time(sec)
Once the quadcopter started to adjust the stability after the motor failure, the unbalance
moments caused most amount of change in yaw angle. As it can be seen, the yaw angle started
92
0.1
φ(rad)
0
-0.1
0 10 20 30 40 50 60
time(sec)
0.02
θ(rad)
-0.02
0 10 20 30 40 50 60
time(sec)
0.4
ψ(rad)
0.2
0
0 10 20 30 40 50 60
time(sec)
Figure (7-18) shows the orientation of the vehicle before and after motor failure.
1.06
θ 4 (rad)
1.05
1.04
0 10 20 30 40 50 60
time(sec)
×10 -3
10
t(rad/s)
-5
0 10 20 30 40 50 60
time(sec)
93
7.2 Preliminary Experimental Results
The tilting rotor quadcopter was tested to find out if it was correctly modeled and controlled.
Extensive simulations were carried out to verify the validity of the dynamic model and control
scheme. This section presents the experimental results obtained with the second prototype. The
first experiment is the hovering task in one spot with tilted orientation. This scenario shows the
performance of the controller and highlights the position control with tilted angle. The second
experiment is intended to demonstrate the performance in tracking a simple trajectory while the
orientation keeps the desired value during the flight. In both scenarios, the altitude maintained
In the first experiment, the scenario was started by taking off and hovering in the fixed
altitude and in one spot while the quadcoter tilts by a desired angle. Figure (7-20) shows the
The tilting starts at t = 8s and continues till it reaches the desired angle at t = 12s. The
vehicle hovers at the same spot with 0.4 rad tilt along the pitch direction till t = 17s and starts
to reduce the tilt to the horizontal flight. At t = 25s it starts to tilt in the opposite direction
along the pitch with the same angle and maintains the orientation till t = 33s when it starts to
1.5
Z(m)
0.5
0
3
2 3
1 2
1
0
0
Y(m) -1 -1 X(m)
During the flight, the vehicle not only maintains the position at the same spot, but also keeps
the roll and yaw close to zero. The altitude is also constant at 1.5m during the experiment.
2
Pitch(rad)
-2
0 5 10 15 20 25 30 35 40
time(sec)
2
Roll(rad)
-2
0 5 10 15 20 25 30 35 40
time(sec)
2
Yaw(rad)
-2
0 5 10 15 20 25 30 35 40
time(sec)
Figure 7-22: The actual orientation of the vehicle along the three directions
95
7.2.2 Trajectory
7.2.2.1 Line
Similarly, we carried out another experiment where the UAV performed the following oper-
ation: i) it took off and went vertically up till t = 15s and altitude of 1.75m; ii) then the vehicle
tilted along the roll direction till t = 17s; iii) the vehicle was then commanded to moved hori-
zontally along with Y axes in the tilted position till t = 32s; iv) the vehicle was then commanded
to come back to horizontally aligned orientation (i.e., no tilt) till t = 35s; and finally, v) the
1.5
Z(m)
0.5
0
3
3
2
2
1 1
0 0
Y(m) -1 -1 X(m)
The UAV’s trajectory is shown in Figure (7-23) and the plot of pitch, roll, and yaw angles
versus time are shown in Figure (7-24). It should be noted that the spikes in the plots come
96
2
Pitch(rad)
0
-2
0 5 10 15 20 25 30 35 40
time(sec)
0.5
Roll(rad)
-0.5
0 5 10 15 20 25 30 35 40
time(sec)
2
Yaw(rad)
-2
0 5 10 15 20 25 30 35 40
time(sec)
Figure (7-24) shows that the vehicle maintained the desired roll angle while moving for-
ward. During the flight, the pitch and the yaw angle were supposed to be around zero.
0.8
0.6
X(m)
0.4
0.2
0 5 10 15 20 25 30 35 40
time(sec)
2
1
Y(m)
-1
0 5 10 15 20 25 30 35 40
time(sec)
2
Z(m)
0
0 5 10 15 20 25 30 35 40
time(sec)
Figure 7-25: Vehicle’s position versus time graph along the 3 directions
97
7.2.2.2 Box
Another trajectory were carried out to show that the vehicle can track with any desired
orientation angle in the range of operation. The vehicle was supposed to move in a square
shaped path. The vehicle was commanded to take off and hover in the height of 1.8m in t = 8s
and stay at the same altitude till t = 55s. The vehicle moved along X axis first. During the
flight, the pitch angle was supposed to be maintained by 0.5rad from t = 15s till t = 50s.
1.5
Z(m)
0.5
0
2
-1
-0.5 1
0
0.5 0
1
X(m) Y(m)
1.5 -1
As it can be seen from (7-26), during the flight, the altitude is constant after short transient
take-off period.
98
1
Pitch(rad)
0.5
-0.5
0 10 20 30 40 50 60
time(sec)
1
Roll(rad)
0.5
-0.5
0 10 20 30 40 50 60
time(sec)
1
Yaw(rad)
0.5
-0.5
0 10 20 30 40 50 60
time(sec)
Figure (7-27) shows the orientation of the vehicle during the flight. The vehicle was com-
1
X(m)
-1
0 10 20 30 40 50 60
time(sec)
2
1
Y(m)
-1
0 10 20 30 40 50 60
time(sec)
2
Z(m)
0
0 10 20 30 40 50 60
time(sec)
Figure 7-28: Vehicle’s position versus time graph along the 3 directions
99
During this period of the time, regardless of direction of the vehicle along X or Y axis,
the pitch was commanded to maintain at 0.5 rad. Figure (7-28) clearly shows the vehicle’s
100
Chapter 8
8.1 Conclusions
In this work, the dynamic modeling and control of a tilting rotor quadcopter was presented.
The tilting rotor design was accomplished by having the tilting capability for the four rotors
achieved via four independent motors used to tilt the rotors along their respective tilting axes.
The relationship between the tilting-rotor angles and the quadcopter orientation was derived
using the dynamic model. However, due to added complexity arising from nonlinear relations
with four additional inputs, and the correlation between these inputs and forces, the Newton-
Euler equation of motions become highly nonlinear. The first two theorems were defined in
order to derive simple relationships between the angular rotations along the tilt directions and
the orientation of the vehicle and rotational speeds of the propellers for hovering. By definition
of these theorems, it was shown that this design makes the quadcopter a fully-controlled system
Hovering with controlled pitch and roll angles, and motion with desired orientation are
This work, furthermore, suggests a PD based method to control the rotational speeds of the
motors responsible for rotating as well as tilting the rotors in order to follow a desired trajectory.
to track a trajectory with a desired orientation during the flight. As the behavior of the tilting
quadcopter, affected by the tilting forces and moments, is highly nonlinear, the linearization
turns out to be a proper control technique to avoid complex behavior of the dynamics. How-
101
ever, in the presence of unmodeled dynamics or any other undefined disturbances, the system
still remains nonlinear after feedback linearization loop. Theoretical results of this control
The dynamic modeling and control of a tilting rotor quadcopter in the case of one motor
failure was also been presented. The relationship between the tilting-rotor angles and the quad-
copter orientation was derived using the dynamic model. It was shown that the quadcopter can
be stabilized if the rotor diagonally opposite to the failed rotor is tilted by an angle that has
been calculated by the Euler equations. In order to maintain an altitude to the closest possible
altitude after the failure, the rotational speed of each individual motor has also been calculated.
The paper presents the dynamic model, and suggests a PD based control in order to avoid crash-
ing of quadcopter and continue the mission. The model and the controller are verified with the
Two different vehicles with tilting mechanism were designed and fabricated as a prototype
of the concept. The proposed control scheme was implemented on the prototype during labo-
ratory experiments which demonstrated the capability of the the vehicle to hover and navigate
with tilted orientation. Furthermore, such over-actuated systems promise to provide mecha-
nisms to not only overcome wind disturbances more effectively but also provide tolerance to
The work presented here addresses the dynamic modeling and control of a tilting rotor
quadcopter with design and fabrication of two different prototypes. Future work should be
• Future work could focus on developing modeling and the control techniques when all
rotors can tilt around their axes independently in order to make the flight smoother and
• Based on our experience with the second prototype, which is lighter and more practically
applicable than the first one, and also by the experience our group had in the other projects
102
with 3D printed quadcopter frame, a third generation prototype of the quadcopter with
tilting rotors can be a fabricated by a combination of 3D printed parts and the carbon fiber
bances in outdoor application more effectively. Further work is needed to develop control
• In order to decrease the reaction time of the tilting mechanism, actuation system improve-
ment as well as higher computational processor and better on-board sensors can be inves-
tigated.
• Further studies are required for understanding the time delay between the time of motor
failure and when the new controller takes over the system. Furthermore, this logic can be
• Implement nonlinear control theory to allow the prototype with one motor failure, to track
complex trajectory.
• Experimental work for executing autonomous flight with tilting rotors also forms a future
work. This would require developing auto-pilot from scratch since available auto-pilot
systems are developed for traditional quadcopters that attempt to modify rotor speeds to
make the frame horizontal when they receive IMU feedback that quad is tilted. Some
progress in this respect has been made. The commands which are required to be sent to
the servo motors with respect to the output of designed controller in this work has been
modified in the code in two different auto pilot boards for the tilt rotor quadcopter. For
the experiment related to the motor failure, the KK2.1 flight controller has been used. The
experiments were done by taking off while the quadcopter had three working motors. This
experiment can be performed as a future work in the way that starts with the quadcopter
with all working motors. Subsequently, fault is induced in one of the motors mid-flight
that requires switching to the failure mode control algorithm in an online fashion. This
experiment would ensure that the transition between different control algorithms is smooth
[2] J. Allen and B. Walsh. Enhanced oil spill surveillance, detection and monitoring through
the applied technology of unmanned air systems. In International oil spill conference,
kalman filter for actuator fault detection and diagnosis of an unmanned quadrotor heli-
[4] D. Andrea. Battery Management Systems for Large Lithium Ion Battery Packs. Artech
house, 2010.
2005.
[6] L. Besnard, Y. B. Shtessel, and B. Landrum. Quadrotor vehicle control via sliding mode
controller driven by sliding mode disturbance observer. Journal of the Franklin Institute,
349(2):658–684, 2012.
104
[8] M. Blanke, M. Staroswiecki, and N. E. Wu. Concepts and methods in fault-tolerant con-
trol. In American Control Conference, 2001. Proceedings of the 2001, volume 4, pages
[9] S. Bouabdallah, P. Murrieri, and R. Siegwart. Towards autonomous indoor micro vtol.
fire surveillance using a team of small unmanned air vehicles. International Journal of
[12] A. Cavoukian. Privacy and drones: Unmanned aerial vehicles. Information and Privacy
[13] E. Cetinsoy. Design and control of a gas-electric hybrid quad tilt-rotor uav with morphing
wing. In Unmanned Aircraft Systems (ICUAS), 2015 International Conference on, pages
[14] H. Chao, Y. Cao, and Y. Chen. Autopilots for small unmanned aerial vehicles: a survey.
Nasa/army/bell xv-15 tiltrotor low noise terminal area operations flight research program.
[16] S. Cousins. Exponential growth of ros [ros topics]. IEEE Robotics & Automation Maga-
[17] P. Doherty and P. Rudol. A uav search and rescue scenario with human body detection
2007.
[18] C. P. Ellington. The novel aerodynamics of insect flight: applications to micro-air vehi-
wireless sensor networks cooperating with uavs for mission-critical management. Wire-
[20] A. Freddi, S. Longhi, A. Monteriu, and M. Prist. Actuator fault detection and isola-
tion system for an hexacopter. In Mechatronic and Embedded Systems and Applications
(MESA), 2014 IEEE/ASME 10th International Conference on, pages 1–6. IEEE, 2014.
[21] S. Griffiths, J. Saunders, A. Curtis, B. Barber, T. McLain, and R. Beard. Obstacle and
terrain avoidance for miniature aerial vehicles. In Advances in Unmanned Aerial Vehicles,
[22] S. Grzonka, G. Grisetti, and W. Burgard. A fully autonomous indoor quadrotor. Robotics,
[23] C. Hancer, K. T. Oner, E. Sirimoglu, E. Cetinsoy, and M. Unel. Robust position control
of a tilt-wing quadrotor. In Decision and Control (CDC), 2010 49th IEEE Conference on,
[24] K. hmadi, A. Y. Javaid, and E. Salari. An efficient compression scheme based on adaptive
flight dynamics and control: Theory and experiment. In Proc. of the AIAA Guidance,
[26] M.-D. Hua, T. Hamel, and C. Samson. Control of vtol vehicles with thrust-tilting aug-
[27] A. S. Huang, E. Olson, and D. C. Moore. Lcm: Lightweight communications and mar-
shalling. In Intelligent robots and systems (IROS), 2010 IEEE/RSJ international confer-
106
[28] V. Hugel, A. Abourachid, H. Gioanni, L. Mederreg, M. Maurice, O. Stasse, P. Bonnin,
and P. Blazevic. The robocoq project: Modelling and design of a bird-like robot equipped
2005.
[31] J. G. Kassakian, H.-C. Wolf, J. M. Miller, and C. J. Hurton. Automotive electrical systems
[32] H. K. Khalil and J. Grizzle. Nonlinear systems, volume 3. Prentice hall New Jersey, 1996.
[34] A. Ö. Kivrak. Design of control systems for a quadrotor flight vehicle equipped with
[35] L.-C. Lai, C.-C. Yang, and C.-J. Wu. Time-optimal control of a hovering quad-rotor
[36] D. Lee, H. J. Kim, and S. Sastry. Feedback linearization vs. adaptive sliding mode control
7(3):419–428, 2009.
[37] D.-J. Lee, I. Kaminer, V. Dobrokhodov, and K. Jones. Autonomous feature following for
visual surveillance using a small unmanned aerial vehicle with gimbaled camera system.
[38] V. Lippiello, F. Ruggiero, and D. Serra. Emergency landing for a quadrotor in case of a
propeller failure: A pid based approach. In Safety, Security, and Rescue Robotics (SSRR),
107
[39] Y. Liu, Z. Pan, D. Stirling, and F. Naghdy. Control of autonomous airship. In Robotics
and Biomimetics (ROBIO), 2009 IEEE International Conference on, pages 2457–2462.
IEEE, 2009.
[40] L. Ljung. System identification toolbox. The MathWorks Inc., South Natick, MA, USA,
1988.
[41] R. Luxman and X. Liu. Implementation of back-stepping integral controller for a gesture
driven quad-copter with human detection and auto follow feature. In Computer Science,
Computer Engineering, and Social Media (CSCESM), 2015 Second International Con-
[42] M. Mahmoud, J. Jiang, and Y. Zhang. Active fault tolerant control systems: stochastic
analysis and synthesis, volume 287. Springer Science & Business Media, 2003.
[43] M. D. Maisel, D. J. Giulianetti, and D. C. Dugan. The history of the xv-15 tilt rotor
[44] L. Meier, P. Tanskanen, F. Fraundorfer, and M. Pollefeys. Pixhawk: A system for au-
tonomous flight using onboard computer vision. In Robotics and automation (ICRA),
A micro aerial vehicle design for autonomous flight using onboard computer vision. Au-
[46] N. Michael, D. Mellinger, Q. Lindsey, and V. Kumar. The grasp multiple micro-uav
[47] T. Mikami and K. Uchiyama. Design of flight control system for quad tilt-wing uav. In
Unmanned Aircraft Systems (ICUAS), 2015 International Conference on, pages 801–805.
IEEE, 2015.
[48] M. W. Mueller and R. D’Andrea. Stability and control of a quadrocopter despite the
108
complete loss of one, two, or three propellers. In Robotics and Automation (ICRA), 2014
[49] D. R. Nelson, D. B. Barber, T. W. McLain, and R. W. Beard. Vector field path following
for miniature air vehicles. Robotics, IEEE Transactions on, 23(3):519–529, 2007.
[50] A. Nemati, M. HZarif, and M. M. Fateh. Helicopter adaptive control with parameter es-
[51] A. Nemati and M. Kumar. Modeling and control of a single axis tilting quadcopter. In
[52] A. Nemati and M. Kumar. Non-linear control of tilting-quadcopter using feedback lin-
earization based motion control. In ASME 2014 Dynamic Systems and Control Confer-
2014.
[54] A. Nemati, R. Kumar, and M. Kumar. Stabilizing and control of tilting-rotor quadcopter
in case of a propeller failure. In ASME 2016 Dynamic systems and control conference.
[55] A. Nemati, N. Soni, M. Sarim, and M. Kumar. Design, fabrication and control of a tilt
rotor quadcopter: Theory and experiments. In ASME 2016 Dynamic systems and control
Robots: Unmanned Aerial Vehicles and Micro Aerial Vehicles. Springer Science & Busi-
[57] B. Norton. Bell Boeing V-22 Osprey: tiltrotor tactical transport. Aerofax, 2004.
model and control of a new quadrotor unmanned aerial vehicle with tilt-wing mechanism.
[60] W. R. Oney. High power density brushless dc motor, Feb. 5 1980. US Patent 4,187,441.
[61] A. Oosedo, S. Abiko, S. Narasaki, A. Kuno, A. Konno, and M. Uchiyama. Flight con-
trol systems of a quad tilt rotor unmanned aerial vehicle for a large attitude change. In
Robotics and Automation (ICRA), 2015 IEEE International Conference on, pages 2326–
[62] C. Papachristos, K. Alexis, and A. Tzes. Hybrid model predictive flight mode conversion
[63] P. Pounds and R. Mahony. Design principles of large quadrotors for practical applications.
In Robotics and Automation, 2009. ICRA’09. IEEE International Conference on, pages
[64] P. Pounds, R. Mahony, and P. Corke. Modelling and control of a quad-rotor robot. In Pro-
[65] L. Qing, C. Zhihao, Y. Jinpeng, S. Yun, and W. YingXun. Trajectory tracking control for
hovering and acceleration maneuver of quad tilt rotor uav. In Control Conference (CCC),
[66] M. Quigley, K. Conley, B. Gerkey, J. Faust, T. Foote, J. Leibs, R. Wheeler, and A. Y. Ng.
Ros: an open-source robot operating system. In ICRA workshop on open source software,
[67] M. Radmanesh and M. Kumar. Grey wolf optimization based sense and avoid algorithm
for UAV path planning in uncertain environment using a bayesian framework. In 2016
2016.
110
[68] M. Radmanesh, M. Kumar, A. Nemati, and M. Sarim. Dynamic optimal UAV trajectory
planning in the national airspace system via mixed integer linear programming. Proceed-
[70] M. Radmanesh, M. Kumar, and M. Sarim. On the effect of different splines on way-point
presence of dynamic obstacles using mixed integer linear programming. In ASME 2015
[72] M. Radmanesh and I. Samani. IUT MAV2013, Part I: Aerodynamic design of tailless
[73] M. Radmanesh and I. Samani. IUT MAV2013, Part II: flight test results.
[74] A. Rodić and G. Mester. The modeling and simulation of an autonomous quad-rotor
2011.
[75] J. Roldan, D. Sanz, J. del Cerro, and A. Barrientos. Lift failure detection and management
system for quadrotors. In ROBOT2013: First Iberian Robotics Conference, pages 103–
[76] M. Ryll, H. Bulthoff, and P. R. Giordano. Modeling and control of a quadrotor uav
with tilting propellers. In Robotics and Automation (ICRA), 2012 IEEE International
111
[77] M. Ryll, H. H. Bulthoff, and P. R. Giordano. First flight tests for a quadrotor uav with tilt-
ing propellers. In Robotics and Automation (ICRA), 2013 IEEE International Conference
[79] E. Samiei, M. Nazari, E. Butcher, and H. Schaub. Delayed feedback control of rigid
space objects from visual feedback during proximity operations. In AIAA/AAS Astrody-
[81] P. Segui-Gasco, Y. Al-Rihani, H.-S. Shin, and A. Savvaris. A novel actuation concept for
a multi rotor uav. Journal of Intelligent & Robotic Systems, 74(1-2):173–191, 2014.
[82] F. Senkul and E. Altug. Modeling and control of a novel tilt-roll rotor quadrotor uav.
In Unmanned Aircraft Systems (ICUAS), 2013 International Conference on, pages 1071–
[83] S. Sepasi, R. Ghorbani, and B. Y. Liaw. Soc estimation for aged lithium-ion batteries using
[84] S. Sepasi, R. Ghorbani, and B. Y. Liaw. Improved extended kalman filter for state of
[85] S. Shafiei and S. Sepasi. Incorporating sliding mode and fuzzy controller with bounded
104(8):3–8, 2015.
112
[86] J.-J. E. Slotine, W. Li, et al. Applied nonlinear control, volume 199. Prentice-Hall Engle-
[87] S. H. Strogatz. Nonlinear dynamics and chaos: with applications to physics, biology,
[88] E. Stura and C. Nicolini. New nanomaterials for light weight lithium batteries. Analytica
[89] A. Tayebi and S. McGilvray. Attitude stabilization of a vtol quadrotor aircraft. IEEE
[91] M. N. Thompson. Flight direction control system for blimps, May 25 1999. US Patent
5,906,335.
[92] L. A. Zadeh and C. A. Deoser. Linear system theory. Robert E. Krieger Publishing
Company, 1976.
[94] Q.-L. Zhou, Y. Zhang, C.-A. Rabbath, and D. Theilliol. Design of feedback lineariza-
tion control and reconfigurable control allocation with application to a quadrotor uav. In
Control and Fault-Tolerant Systems (SysTol), 2010 Conference on, pages 371–376. IEEE,
2010.
113